Quick Summary
Goal: Produce a 3-way estimation calibration report — pre_impl_estimate (from plan) vs true_estimate (from observed scope) vs actual_time (from git/user) — yielding two INDEPENDENT signals: developer execution variance and estimation model calibration variance.
Why two signals matter: They are confounded if not separated. If actual >> pre-impl, the bug could be (a) developer was slow, OR (b) the model under-estimated scope. Without computing TRUE from observed scope, you cannot tell which. Single-sample calibration has near-zero statistical power — the skill always reports this.
Workflow:
- Detect input mode — plan-file path /
--changes / --pr <n>
- Read pre-impl estimate from plan frontmatter (if present)
- Observe actual scope — git diff, file list, line counts, blast radius from the diff
- Compute TRUE estimate — apply the canonical estimation framework (inlined below) to the OBSERVED scope (post-hoc, with full visibility)
- Get actual time — derive from git timestamps if available; ask user to confirm/override (timestamps ≠ coding time)
- Compute variances — scope variance (TRUE vs pre-impl) and execution variance (actual vs TRUE)
- Per-layer breakdown — UI tier, backend tier, test count, blast radius — predicted vs observed
- Report — calibration verdict + suggested model adjustments (only if pattern across ≥3 samples)
Key Rules:
- MUST ATTENTION compute TRUE estimate using the SAME canonical framework that was used for pre-impl — fair comparison requires identical methodology
- MUST ATTENTION separate developer execution signal from model calibration signal — never collapse to a single "good/bad estimate" verdict
- NEVER claim a model adjustment from a single sample — explicitly state "needs ≥3 samples for signal"
- NEVER trust git timestamps as actual coding time — they include sleep, meetings, context switches; ALWAYS ask user to validate
- MUST ATTENTION list per-layer deltas (UI/backend/tests/blast) — aggregate variance hides where the model went wrong
- Use min-max range for both pre-impl and TRUE — comparing single points is dishonest about uncertainty
Input Modes
| Mode |
Trigger |
What's read |
| Plan-file |
<path/to/plan.md> |
Plan frontmatter (pre-impl estimate) + git diff scoped to plan branch |
| Changes |
--changes |
git diff working tree + last commit timestamps |
| PR |
--pr <n> |
gh pr view <n> + gh pr diff <n> + PR open/merge times |
If multiple modes detected (e.g., plan file AND --changes), prefer plan-file (carries the original estimate); use changes for the diff source.
Workflow Detail
Step 1: Detect mode and gather pre-impl estimate
- Plan-file mode: read frontmatter, extract
man_days_traditional, story_points, risk_margin_pct, risk_factors, blast_radius, estimate_reasoning if present
- Changes/PR mode without plan: skip pre-impl; report only TRUE vs actual
- If plan exists but uses old single-point format (no range, no
risk_factors), flag in report — comparison is approximate
Step 2: Observe actual scope from diff
Run (PowerShell or Bash via tool):
git diff --stat <base>..<head> — file count, lines added/removedgit diff --name-only <base>..<head> — file listgit log --format='%H %ai %s' <base>..<head> — commit timeline
Classify changed files:
- UI files (component/template/style) — count, group by screen
- Backend handlers/entities/repos — count, classify per backend tier table
- Tests (unit/integration/e2e) — count test files, count test cases (grep
describe|it|Fact|Test\b)
- Migrations / contracts / shared code — flag separately
Step 3: Run Blast Radius pass on observed scope
- Touched files / components — count
- Of those, complex (>500 LOC area, multi-handler, central) — count
- Downstream consumers — use code graph trace if available:
python .claude/scripts/code_graph trace <file> --direction both --json for changed entry-point files
- Shared/common code touched — yes/no
- Regression scope — list affected areas
Step 4: Apply canonical estimation framework to observed scope
Apply each tier table (UI / backend / test / risk margin / risk factors) from the inline framework below to the OBSERVED scope. Output:
true_likely_days (single midpoint)true_min_days = likely × 0.9true_max_days = likely × (1 + risk_margin)true_estimate = '<min>-<max>d' range
Step 5: Get actual time
Try in order:
- Git: timestamp of first commit on feature branch → timestamp of last commit (or merge commit). Convert to working days (8h business days, exclude weekends).
- PR: open time → merge time. Same conversion.
- Ask user via
AskUserQuestion: "Git suggests N working days from first commit to merge. How much was actual coding time? (excludes meetings, code-review wait, context switches, vacations)"
ALWAYS surface the gap between elapsed time and reported coding time — they are different signals.
Step 6: Compute variances
scope_variance_pct = (true_likely - preimpl_likely) / preimpl_likely × 100
exec_variance_pct = (actual_time - true_likely) / true_likely × 100
Interpretation matrix:
| scope_var |
exec_var |
Verdict |
| ~0% (±15%) |
~0% (±15%) |
Estimate matched scope; developer matched estimate. Healthy. |
| ~0% |
>+25% |
Model OK; developer slower than expected. Performance signal. |
| ~0% |
<-25% |
Model OK; developer faster than expected. Either skilled or scope simpler than apparent. |
| >+25% |
~0% |
Model UNDER-estimated scope; developer matched the harder-than-predicted reality. Model signal — too optimistic. |
| <-25% |
~0% |
Model OVER-estimated scope; actual work was simpler. Model signal — too pessimistic. |
| >+25% |
>+25% |
Both — scope was harder AND developer slower. Disambiguate over multiple samples. |
| <-25% |
<-25% |
Original estimate was way over; developer also fast. Likely simple task padded heavily. |
Step 7: Per-layer breakdown (where the model went wrong)
| Layer |
Pre-impl tier (from plan) |
Observed tier (from diff) |
Delta |
| UI |
e.g. "Compose components into NEW screen" |
e.g. "Add control to existing screen" |
-1 tier (~0.7d over) |
| Backend |
e.g. "NEW command on existing aggregate" |
e.g. "Small update existing handler" |
-1 tier (~0.5d over) |
| Tests |
e.g. "13 cases" |
e.g. "5 cases" |
-8 cases (~0.5d over) |
| Blast |
e.g. "4 areas, 1 complex" |
e.g. "2 areas, 0 complex" |
lower regression risk |
| Risk factors |
predicted list |
applicable in retrospect |
call out missing/unused |
Step 8: Report
Produce a markdown report with sections:
- Summary table — three numbers (pre-impl range, TRUE range, actual single)
- Variance verdict — interpretation matrix row + plain-English explanation
- Per-layer breakdown — table above
- Risk factors — predicted vs applicable; note any new factors that surfaced (e.g., regression-fan-out not flagged but should have been)
- Calibration suggestion — ONLY if user has run this skill ≥3 times with consistent direction. Single-sample → state "no statistical power, log this sample for future calibration"
- Confidence — state confidence level for each verdict; uncertainty about actual time goes here
Step 9: Persist sample (optional)
If user wants longitudinal tracking, append the calibration row to plans/_estimation-samples.csv:
date,plan,preimpl_min,preimpl_max,true_min,true_max,actual,scope_var_pct,exec_var_pct,risk_factors_predicted,risk_factors_applicable
After ≥5 rows, run pattern detection on the CSV: if scope_var_pct is consistently negative (model over-estimates), suggest tier adjustment; if consistently positive (under-estimates), suggest adding risk factors or widening tier.
Estimation Framework (canonical — applied in Step 4)
The canonical framework lives in the Estimation Framework sync block at the end of this skill; Step 4 applies it verbatim to the observed (post-hoc) scope.
Output Report Template
# Estimation Calibration Report — <plan or branch name>
## Summary
| Metric | Range / Value | Source |
| ----------------- | -------------------------- | ---------------------------------------- |
| Pre-impl estimate | <min>-<max>d (likely <m>d) | <plan path frontmatter> |
| TRUE estimate | <min>-<max>d (likely <m>d) | observed scope (post-hoc) |
| Actual time | <n>d | git <first commit→merge>, user-confirmed |
**Scope variance** (TRUE vs pre-impl): <±n>% — <under/over/matched>
**Execution variance** (actual vs TRUE likely): <±n>% — <fast/slow/matched>
## Verdict
| Signal | Direction | Magnitude | Confidence |
| ------------------- | ----------------------------------------------- | --------- | ----------------- |
| Estimation model | <too optimistic / too pessimistic / calibrated> | <±n>% | <low/medium/high> |
| Developer execution | <fast / slow / on-pace> | <±n>% | <low/medium/high> |
## Per-Layer Breakdown
| Layer | Predicted tier | Observed tier | Delta |
| ------------ | ------------------ | ------------------ | --------------- |
| UI | … | … | … |
| Backend | … | … | … |
| Tests | … cases | … cases | … |
| Blast radius | … areas, … complex | … areas, … complex | … |
| Risk factors | <predicted list> | <applicable list> | <added/removed> |
## Calibration Suggestions
- <If single sample> No model adjustment from one data point. Logged to `plans/_estimation-samples.csv` (row N). Re-run /estimate-actual on future plans to build calibration corpus. Suggested adjustment after ≥3-5 samples with consistent direction.
- <If pattern across samples> e.g. "UI tier 'Compose components into NEW screen' overshoots in 4/5 samples by ~0.5d → suggest splitting into two tiers OR widening band to 1-2.5d"
## Caveats
- Actual time derived from <git/user>; <list any uncertainty: weekends, code-review days, vacations excluded?>
- Pre-impl estimate format <range/single-point/missing> — comparison <exact/approximate>
- Confidence in TRUE estimate: <high/medium/low> — observed scope <fully visible / partially obscured>
Anti-Rationalization Anchors
| Evasion |
Rebuttal |
| "Single sample is enough — clearly the dev was slow" |
NO. Without separating scope from execution variance, you confound model error and performance. State signal + caveat. |
| "Use git timestamps as actual time" |
Wrong. Includes weekends, meetings, code-review wait, sleep. Always confirm with user. |
| "Skip TRUE estimate — just compare pre-impl vs actual" |
That's the data point that's MISSING and exactly why estimates don't improve over time. Never skip Step 4. |
| "Apply hindsight to pump up TRUE estimate" |
Use the SAME framework that was used for pre-impl. Hindsight bias inflates TRUE and falsely vindicates the original estimate. |
| "One signal is fine, no need to split" |
Two signals is the entire point. Performance review needs execution variance; model tuning needs scope variance. Confounded data is unactionable. |
Estimation Framework — Bottom-up first; SP DERIVED; output min-max range when likely ≥3d. Stack-agnostic. Baseline: 3-5yr dev, 6 productive hrs/day. AI estimate assumes Claude Code + project context.
Method:
- Blast Radius pass (below) — drives code AND test cost
- Decompose phases → hours/phase →
bottom_up_hours = Σ phase_hours
likely_days = ceil(bottom_up_hours / 6) × productivity_factor- Sum Risk Margin (base + add-ons) →
max_days = likely_days × (1 + margin)
min_days = likely_days × 0.9- Output as range when
likely_days ≥3; single point allowed <3 (still record margin)
man_days_ai = same range × AI speedupstory_points DERIVED from likely_days via SP-Days — NEVER driver. Disagreement >50% → trust bottom-up
Productivity factor: 0.8 strong scaffolding+codegen+AI hooks · 1.0 mature default · 1.2 weak patterns · 1.5 greenfield
Cost Driver Heuristic (apply BEFORE work-type row):
- UI dominates in CRUD/business apps — 1.5-3x backend (states, validation, responsive, a11y, polish)
- Backend dominates ONLY: multi-aggregate invariants, cross-service contracts, schema migrations, heavy query/perf, new event flows
Reuse-vs-Create axis (PRIMARY lever, per layer):
| UI tier |
Cost |
| Reuse component on existing screen |
0.1-0.3d |
| Add control/column to existing screen |
0.3-0.8d |
| Compose components into NEW screen |
1-2d |
| NEW screen, custom layout/states/validation |
2-4d |
| NEW shared/common component (themed, tested) |
3-6d+ |
| Backend tier |
Cost |
| Reuse query/handler from new place |
0.1-0.3d |
| Small update existing handler/entity |
0.3-0.8d |
| NEW query on existing repo/model |
0.5-1d |
| NEW command/handler on existing aggregate (additive) |
1-2d |
| NEW aggregate/entity (repo, validation, events) |
2-4d |
| NEW cross-service contract OR schema migration |
2-4d each |
| Multi-aggregate invariant / heavy domain rule |
3-5d |
Rule: Sum tiers across UI+backend+tests, apply productivity factor. Reuse short-circuits tiers — call out.
Test-Scope drivers (compute test_count EXPLICITLY — "+tests" hand-wave is #1 failure):
| Driver |
Count |
| Happy-path journeys |
1 per story / AC main flow |
| State-machine transitions |
reachable transitions × allowed actors |
| Multi-entity state combos |
state(A) × state(B) — REACHABLE only, not Cartesian |
| Authorization matrix |
(owner, non-owner, elevated, unauth) × each mutation |
| Validation rules |
1 per required field / boundary / format / cross-field |
| UI states (per new screen/dialog) |
happy, loading, empty, error, partial — present only |
| Negative paths / invariants |
1 per violatable business rule |
| Test tier (Trad, incl. setup+assert+flake) |
Cost |
| 1-5 cases, fixtures reused |
0.3-0.5d |
| 6-12 cases, 1 new fixture |
0.5-1d |
| 13-25 cases, multi-entity setup |
1-2d |
| 26-50 cases OR new state-machine coverage |
2-3d |
| >50 cases OR full E2E journey |
3-5d |
Test multipliers: new fixture/seed harness +0.5d · cross-service/bus assertion +0.3d each · UI E2E ×1.5 · each new role +1-2 cases
Blast Radius (mandatory pre-pass — affects code AND test):
- Files/components directly modified — count
- Of those, "complex" (>500 LOC, multi-handler, central, frequently-modified) — count
- Downstream consumers (callers, event subscribers, cross-service) — list
- Shared/common code touched (multi-app blast) — yes/no
- Regression scope — areas needing re-test
Rule: Complex touch → add risk_factors. Each downstream consumer → +1-3 regression cases. Blast >5 areas OR >2 complex → re-evaluate SPLIT before estimating.
Risk Margin (drives max bound):
| likely_days |
Base margin |
| <1d trivial |
+10% |
| 1-2d small additive |
+20% |
| 3-4d real feature |
+35% |
| 5-7d large |
+50% |
| 8-10d very large |
+75% |
| >10d |
+100% AND flag SHOULD SPLIT |
Risk-factor add-ons (additive — enumerate in risk_factors):
| Factor |
+margin |
touches-complex-existing-feature (>500 LOC, multi-handler, central) |
+20% |
cross-service-contract change |
+25% |
schema-migration-on-populated-data |
+25% |
new-tech-or-unfamiliar-pattern |
+30% |
regression-fan-out (≥3 downstream areas re-test) |
+20% |
performance-or-latency-critical |
+20% |
concurrency-race-event-ordering |
+25% |
shared-common-code (multi-consumer/multi-app) |
+25% |
unclear-requirements-or-design |
+30% |
Collapse rule: total margin >100% → STOP, split (padding past 2x is dishonesty). Margin <15% on likely_days ≥5 → under-estimated, widen.
Work-Type Caps (hard ceilings on likely_days):
| Work type |
Max SP |
Max likely |
| Single field / config flag / style fix |
1 |
0.5d |
| Add property to existing model + bind to existing UI |
2 |
1d |
| Additive endpoint + minor UI control (button/menu/column), reuses fixtures |
3 |
2-3d |
| Additive endpoint + NEW UI surface OR additive multi-layer + new domain rule + 2+ test files |
5 |
3-5d |
| NEW model/aggregate OR migration OR cross-module contract OR heavy test (>1.5d) OR NEW UI + non-trivial backend |
8 |
5-7d |
| NEW UI surface + (NEW aggregate OR migration OR cross-service contract) |
13 |
SHOULD split |
| Cross-service contract + migration combined |
13 |
SHOULD split |
| Beyond |
21 |
MUST split |
SP→Days (validation only): 1=0.5d/0.25d · 2=1d/0.35d · 3=2d/0.65d · 5=4d/1.0d · 8=6d/1.5d · 13=10d/2.0d (Trad/AI likely)
AI speedup: SP 1≈2x · 2-3≈3x · 5-8≈4x · 13+≈5x. AI cost = (code_gen × 1.3) + (test_gen × 1.3) (30% review overhead).
MANDATORY frontmatter:
story_points: <n>
complexity: low | medium | high | critical
man_days_traditional: '<min>-<max>d' # range when likely ≥3d; '<N>d' when <3d
man_days_ai: '<min>-<max>d'
risk_margin_pct: <n> # base + add-ons
risk_factors: [touches-complex-existing-feature, regression-fan-out] # closed-list from add-ons; [] if none
blast_radius:
touched_areas: <n>
complex_touched: <n>
downstream_consumers: [list or count]
shared_common_code: yes | no
estimate_scope_included: [code, integration-tests, frontend, i18n, docs]
estimate_scope_excluded: [unit-tests, e2e, perf, deployment, code-review-rounds]
estimate_reasoning: |
5-7 lines covering:
(a) UI tier — row applied
(b) Backend tier — row applied
(c) Test scope — case breakdown by driver, file count, fixtures, tier row
(d) Cost driver — dominant tier + why
(e) Blast radius — touched, complex, regression scope
(f) Risk factors — list driving margin; why not larger/smaller
Example: "UI: compose Form/Table/Dialog → NEW screen (~1.5d). Backend: NEW command on existing aggregate,
reuses validation+repo (~1d). Tests: 4 transitions × 2 actors + 3 validation + 2 UI states = 13 cases,
1 new fixture → tier 13-25 ~1.5d. Driver: UI composition + new states. Blast: 4 areas, 1 complex.
Risk: base 35% + touches-complex +20% = 55% → max 3.9d → range 2.5-4d."
Sanity self-check:
likely_days ≥3d and single-point? → reject, must be range- Margin <15% on
likely_days ≥5d? → under-estimated, widen
- Margin >100%? → STOP, split instead of buffer
- Complex existing feature touched, no regression budget in
(c)? → reject
- Blast
>5 areas OR >2 complex, no split discussion? → reject
- Purely additive on existing model AND existing UI? → cap SP 3 unless tests >1.5d
- NEW UI surface (page/complex form/dashboard)? → SP 5+ even if backend one endpoint
- Backend cross-service / migration / multi-aggregate? → SP 8+ regardless of UI
bottom_up_hours / 6 vs SP-Days disagreement >50%? → trust bottom-up, downgrade SP- Without tests, SP drops ≥1 bucket? → tests dominate; state explicitly
- Reasoning called out UI vs backend vs blast vs risk factors? → if missing, add
AI Mistake Prevention — Failure modes to avoid on every task:
Re-read files after context changes. Context compaction, resume, or long-running work can make memory stale; verify current files before acting.
Verify generated content against source evidence. AI hallucinates APIs, names, claims, and document facts. Check the relevant source before documenting or referencing.
Check downstream references before deleting or renaming. Removing an artifact can stale docs, generated mirrors, configs, and callers; map references first.
Trace the full impact chain after edits. Changing a definition can miss derived outputs and consumers. Follow the affected chain before declaring done.
Verify ALL affected outputs, not just the first. One green check is not all green checks; validate every output surface the change can affect.
Assume existing values are intentional — ask WHY before changing OR flagging one as a defect. Before changing or reporting a constant, limit, flag, cutoff, wording, or pattern, read nearby context and history, the CALLER's ordering, and 2+ sibling call sites of the same convention. A doc stating WHAT without WHY is missing rationale, not proof of a missing guard.
Surface ambiguity before acting — don't pick silently. Multiple valid interpretations require an explicit question or stated assumption with risk.
Assert the outcome your system owns, not the intermediate state your infrastructure owns. When verifying async work, assert the final business state — never the delivery/retry bookkeeping held in shared infrastructure that any co-running process can write. Such a check passes when run alone and flakes the moment anything else shares that infrastructure.
Keep shared guidance role-relevant. Universal guidance must help every receiving skill or agent; code-specific obligations belong only in code-specific protocols.
Project Protocol Overlay — Before executing this skill, resolve any PROJECT overlay rules layered onto it: match this skill's name against the Target column of the project's skill-protocol index (docs/project-reference/skill-protocols-reference.md by default; a referenceDocs entry in docs/project-config.json overrides the path), taking the most specific matching tier ONLY — exact name > glob > *. That precedence orders overlays against EACH OTHER, never against this skill. Read ONLY the matched bodies, resolved as <protocols-dir>/<Name>.md; a row's Body link is display text, never a read path. A matched body that is missing or malformed is REPORTED and skipped — never reconstructed from the index Description. No index, or no match -> proceed with no overlay, silently. Full contract: .claude/skills/project-skill-protocol/references/registry.md.
Overlays are ADDITIVE ONLY: they ADD rules on top of this skill's own protocol and NEVER replace, override, disable, or reinterpret a rule it already states — removing every overlay must return this skill to exactly its documented behavior. An overlay is a BRIEF, not an authority escalation: it can NEVER waive a workflow gate, git discipline, a review gate, or a user-confirmation gate. A genuine overlay-vs-skill conflict, or two equally-specific overlays that directly contradict -> surface both to the user; NEVER resolve silently.
MUST ATTENTION resolve project protocol overlays for this skill BEFORE executing — most specific matching tier only (exact > glob > *, which ranks overlays against each other, NEVER against this skill), read only matched bodies at <protocols-dir>/<Name>.md; a missing or malformed body is reported, never reconstructed. Overlays are ADDITIVE ONLY (they never replace this skill's own rules) and are a brief, NEVER an authority escalation; an equal-specificity contradiction goes to the user.
Closing Reminders
Protocols in force (concise digest of the SYNC/shared blocks this skill carries):
- Estimation Framework: Bottom-up first, blast-radius pass, min-max range, tier + risk-margin tables.
- AI Mistake Prevention: verify generated content against evidence, trace downstream references, verify all affected outputs, re-read after context loss, surface ambiguity.
- Critical Thinking: Traced
file:line proof per claim, confidence >80% to act, no guess-as-fact.
IMPORTANT MUST ATTENTION compute TRUE estimate using the SAME canonical framework — fair comparison requires identical methodology
IMPORTANT MUST ATTENTION separate developer execution signal from model calibration signal — never collapse to single verdict
IMPORTANT MUST ATTENTION never claim model adjustment from a single sample — explicitly state "needs ≥3 samples for signal"
IMPORTANT MUST ATTENTION never trust git timestamps as coding time — always ask user to confirm/override
IMPORTANT MUST ATTENTION list per-layer deltas (UI/backend/tests/blast) — aggregate variance hides where model went wrong
IMPORTANT MUST ATTENTION use min-max ranges for both pre-impl and TRUE — comparing single points is dishonest about uncertainty
IMPORTANT MUST ATTENTION apply Blast Radius pass on observed diff before applying tier tables
IMPORTANT MUST ATTENTION persist samples to plans/_estimation-samples.csv for longitudinal calibration
IMPORTANT MUST ATTENTION state confidence per verdict — uncertainty about actual time goes in caveats
[IMPORTANT] Use TaskCreate to break ALL work into small tasks BEFORE starting.
Critical Thinking Mindset — Apply critical thinking, sequential thinking. Every claim needs traced proof, confidence >80% to act.
Anti-hallucination: Never present guess as fact — cite sources for every claim, admit uncertainty freely, self-check output for errors, cross-reference independently, stay skeptical of own confidence — certainty without evidence root of all hallucination.
MUST ATTENTION apply critical + sequential thinking — every claim needs appropriate traced evidence (file:line for repo/code claims; source URL or artifact section for research, product, content, and docs claims); confidence >80% to act, <60% DO NOT recommend. Anti-hallucination: never present guess as fact, admit uncertainty freely, cross-reference independently, stay skeptical of own confidence.
MUST ATTENTION apply AI mistake prevention — verify generated content against evidence, trace downstream references before deleting or renaming, verify all affected outputs, re-read files after context loss, and surface ambiguity before acting.
[TASK-PLANNING] Before acting, analyze task scope and systematically break it into small todo tasks and sub-tasks using TaskCreate.
1---2name: estimate-actual-33description: [Planning] Use when calibrating estimates from actual code, diff, PR scope, and developer time.4---56## Quick Summary78**Goal:** Produce a 3-way estimation calibration report — `pre_impl_estimate` (from plan) vs `true_estimate` (from observed scope) vs `actual_time` (from git/user) — yielding two INDEPENDENT signals: developer execution variance and estimation model calibration variance.910**Why two signals matter:** They are confounded if not separated. If actual >> pre-impl, the bug could be (a) developer was slow, OR (b) the model under-estimated scope. Without computing TRUE from observed scope, you cannot tell which. Single-sample calibration has near-zero statistical power — the skill always reports this.1112**Workflow:**13141. **Detect input mode** — plan-file path / `--changes` / `--pr <n>`152. **Read pre-impl estimate** from plan frontmatter (if present)163. **Observe actual scope** — git diff, file list, line counts, blast radius from the diff174. **Compute TRUE estimate** — apply the canonical estimation framework (inlined below) to the OBSERVED scope (post-hoc, with full visibility)185. **Get actual time** — derive from git timestamps if available; ask user to confirm/override (timestamps ≠ coding time)196. **Compute variances** — scope variance (TRUE vs pre-impl) and execution variance (actual vs TRUE)207. **Per-layer breakdown** — UI tier, backend tier, test count, blast radius — predicted vs observed218. **Report** — calibration verdict + suggested model adjustments (only if pattern across ≥3 samples)2223**Key Rules:**2425- MUST ATTENTION compute TRUE estimate using the SAME canonical framework that was used for pre-impl — fair comparison requires identical methodology26- MUST ATTENTION separate developer execution signal from model calibration signal — never collapse to a single "good/bad estimate" verdict27- NEVER claim a model adjustment from a single sample — explicitly state "needs ≥3 samples for signal"28- NEVER trust git timestamps as actual coding time — they include sleep, meetings, context switches; ALWAYS ask user to validate29- MUST ATTENTION list per-layer deltas (UI/backend/tests/blast) — aggregate variance hides where the model went wrong30- Use min-max range for both pre-impl and TRUE — comparing single points is dishonest about uncertainty3132## Input Modes3334| Mode | Trigger | What's read |35| --------- | ------------------- | --------------------------------------------------------------------- |36| Plan-file | `<path/to/plan.md>` | Plan frontmatter (pre-impl estimate) + git diff scoped to plan branch |37| Changes | `--changes` | `git diff` working tree + last commit timestamps |38| PR | `--pr <n>` | `gh pr view <n>` + `gh pr diff <n>` + PR open/merge times |3940If multiple modes detected (e.g., plan file AND `--changes`), prefer plan-file (carries the original estimate); use changes for the diff source.4142## Workflow Detail4344### Step 1: Detect mode and gather pre-impl estimate4546- Plan-file mode: read frontmatter, extract `man_days_traditional`, `story_points`, `risk_margin_pct`, `risk_factors`, `blast_radius`, `estimate_reasoning` if present47- Changes/PR mode without plan: skip pre-impl; report only TRUE vs actual48- If plan exists but uses old single-point format (no range, no `risk_factors`), flag in report — comparison is approximate4950### Step 2: Observe actual scope from diff5152Run (PowerShell or Bash via tool):5354- `git diff --stat <base>..<head>` — file count, lines added/removed55- `git diff --name-only <base>..<head>` — file list56- `git log --format='%H %ai %s' <base>..<head>` — commit timeline5758Classify changed files:5960- UI files (component/template/style) — count, group by screen61- Backend handlers/entities/repos — count, classify per backend tier table62- Tests (unit/integration/e2e) — count test files, count test cases (grep `describe|it|Fact|Test\b`)63- Migrations / contracts / shared code — flag separately6465### Step 3: Run Blast Radius pass on observed scope6667- Touched files / components — count68- Of those, complex (>500 LOC area, multi-handler, central) — count69- Downstream consumers — use code graph trace if available: `python .claude/scripts/code_graph trace <file> --direction both --json` for changed entry-point files70- Shared/common code touched — yes/no71- Regression scope — list affected areas7273### Step 4: Apply canonical estimation framework to observed scope7475Apply each tier table (UI / backend / test / risk margin / risk factors) from the inline framework below to the OBSERVED scope. Output:7677- `true_likely_days` (single midpoint)78- `true_min_days = likely × 0.9`79- `true_max_days = likely × (1 + risk_margin)`80- `true_estimate = '<min>-<max>d'` range8182### Step 5: Get actual time8384Try in order:85861. Git: timestamp of first commit on feature branch → timestamp of last commit (or merge commit). Convert to working days (8h business days, exclude weekends).872. PR: open time → merge time. Same conversion.883. Ask user via `AskUserQuestion`: "Git suggests N working days from first commit to merge. How much was actual coding time? (excludes meetings, code-review wait, context switches, vacations)"8990ALWAYS surface the gap between elapsed time and reported coding time — they are different signals.9192### Step 6: Compute variances9394```95scope_variance_pct = (true_likely - preimpl_likely) / preimpl_likely × 10096exec_variance_pct = (actual_time - true_likely) / true_likely × 10097```9899Interpretation matrix:100101| scope_var | exec_var | Verdict |102| ---------- | ---------- | -------------------------------------------------------------------------------------------------------------------- |103| ~0% (±15%) | ~0% (±15%) | Estimate matched scope; developer matched estimate. Healthy. |104| ~0% | >+25% | Model OK; developer slower than expected. **Performance signal.** |105| ~0% | <-25% | Model OK; developer faster than expected. Either skilled or scope simpler than apparent. |106| >+25% | ~0% | Model UNDER-estimated scope; developer matched the harder-than-predicted reality. **Model signal — too optimistic.** |107| <-25% | ~0% | Model OVER-estimated scope; actual work was simpler. **Model signal — too pessimistic.** |108| >+25% | >+25% | Both — scope was harder AND developer slower. Disambiguate over multiple samples. |109| <-25% | <-25% | Original estimate was way over; developer also fast. Likely simple task padded heavily. |110111### Step 7: Per-layer breakdown (where the model went wrong)112113| Layer | Pre-impl tier (from plan) | Observed tier (from diff) | Delta |114| ------------ | ----------------------------------------- | ------------------------------------- | ----------------------- |115| UI | e.g. "Compose components into NEW screen" | e.g. "Add control to existing screen" | -1 tier (~0.7d over) |116| Backend | e.g. "NEW command on existing aggregate" | e.g. "Small update existing handler" | -1 tier (~0.5d over) |117| Tests | e.g. "13 cases" | e.g. "5 cases" | -8 cases (~0.5d over) |118| Blast | e.g. "4 areas, 1 complex" | e.g. "2 areas, 0 complex" | lower regression risk |119| Risk factors | predicted list | applicable in retrospect | call out missing/unused |120121### Step 8: Report122123Produce a markdown report with sections:1241251. **Summary table** — three numbers (pre-impl range, TRUE range, actual single)1262. **Variance verdict** — interpretation matrix row + plain-English explanation1273. **Per-layer breakdown** — table above1284. **Risk factors** — predicted vs applicable; note any new factors that surfaced (e.g., regression-fan-out not flagged but should have been)1295. **Calibration suggestion** — ONLY if user has run this skill ≥3 times with consistent direction. Single-sample → state "no statistical power, log this sample for future calibration"1306. **Confidence** — state confidence level for each verdict; uncertainty about actual time goes here131132### Step 9: Persist sample (optional)133134If user wants longitudinal tracking, append the calibration row to `plans/_estimation-samples.csv`:135136```137date,plan,preimpl_min,preimpl_max,true_min,true_max,actual,scope_var_pct,exec_var_pct,risk_factors_predicted,risk_factors_applicable138```139140After ≥5 rows, run pattern detection on the CSV: if `scope_var_pct` is consistently negative (model over-estimates), suggest tier adjustment; if consistently positive (under-estimates), suggest adding risk factors or widening tier.141142## Estimation Framework (canonical — applied in Step 4)143144The canonical framework lives in the **Estimation Framework** sync block at the end of this skill; Step 4 applies it verbatim to the observed (post-hoc) scope.145146## Output Report Template147148```markdown149# Estimation Calibration Report — <plan or branch name>150151## Summary152153| Metric | Range / Value | Source |154| ----------------- | -------------------------- | ---------------------------------------- |155| Pre-impl estimate | <min>-<max>d (likely <m>d) | <plan path frontmatter> |156| TRUE estimate | <min>-<max>d (likely <m>d) | observed scope (post-hoc) |157| Actual time | <n>d | git <first commit→merge>, user-confirmed |158159**Scope variance** (TRUE vs pre-impl): <±n>% — <under/over/matched>160**Execution variance** (actual vs TRUE likely): <±n>% — <fast/slow/matched>161162## Verdict163164| Signal | Direction | Magnitude | Confidence |165| ------------------- | ----------------------------------------------- | --------- | ----------------- |166| Estimation model | <too optimistic / too pessimistic / calibrated> | <±n>% | <low/medium/high> |167| Developer execution | <fast / slow / on-pace> | <±n>% | <low/medium/high> |168169## Per-Layer Breakdown170171| Layer | Predicted tier | Observed tier | Delta |172| ------------ | ------------------ | ------------------ | --------------- |173| UI | … | … | … |174| Backend | … | … | … |175| Tests | … cases | … cases | … |176| Blast radius | … areas, … complex | … areas, … complex | … |177| Risk factors | <predicted list> | <applicable list> | <added/removed> |178179## Calibration Suggestions180181- <If single sample> No model adjustment from one data point. Logged to `plans/_estimation-samples.csv` (row N). Re-run /estimate-actual on future plans to build calibration corpus. Suggested adjustment after ≥3-5 samples with consistent direction.182- <If pattern across samples> e.g. "UI tier 'Compose components into NEW screen' overshoots in 4/5 samples by ~0.5d → suggest splitting into two tiers OR widening band to 1-2.5d"183184## Caveats185186- Actual time derived from <git/user>; <list any uncertainty: weekends, code-review days, vacations excluded?>187- Pre-impl estimate format <range/single-point/missing> — comparison <exact/approximate>188- Confidence in TRUE estimate: <high/medium/low> — observed scope <fully visible / partially obscured>189```190191## Anti-Rationalization Anchors192193| Evasion | Rebuttal |194| ------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------- |195| "Single sample is enough — clearly the dev was slow" | NO. Without separating scope from execution variance, you confound model error and performance. State signal + caveat. |196| "Use git timestamps as actual time" | Wrong. Includes weekends, meetings, code-review wait, sleep. Always confirm with user. |197| "Skip TRUE estimate — just compare pre-impl vs actual" | That's the data point that's MISSING and exactly why estimates don't improve over time. Never skip Step 4. |198| "Apply hindsight to pump up TRUE estimate" | Use the SAME framework that was used for pre-impl. Hindsight bias inflates TRUE and falsely vindicates the original estimate. |199| "One signal is fine, no need to split" | Two signals is the entire point. Performance review needs execution variance; model tuning needs scope variance. Confounded data is unactionable. |200201---202203<!-- SYNC:estimation-framework -->204205> **Estimation Framework** — Bottom-up first; SP DERIVED; output min-max range when likely ≥3d. Stack-agnostic. Baseline: 3-5yr dev, 6 productive hrs/day. AI estimate assumes Claude Code + project context.206>207> **Method:**208>209> 1. **Blast Radius pass** (below) — drives code AND test cost210> 2. Decompose phases → hours/phase → `bottom_up_hours = Σ phase_hours`211> 3. `likely_days = ceil(bottom_up_hours / 6) × productivity_factor`212> 4. Sum **Risk Margin** (base + add-ons) → `max_days = likely_days × (1 + margin)`213> 5. `min_days = likely_days × 0.9`214> 6. Output as range when `likely_days ≥3`; single point allowed `<3` (still record margin)215> 7. `man_days_ai` = same range × AI speedup216> 8. `story_points` DERIVED from `likely_days` via SP-Days — NEVER driver. Disagreement >50% → trust bottom-up217>218> **Productivity factor:** 0.8 strong scaffolding+codegen+AI hooks · 1.0 mature default · 1.2 weak patterns · 1.5 greenfield219>220> **Cost Driver Heuristic (apply BEFORE work-type row):**221>222> - **UI dominates** in CRUD/business apps — 1.5-3x backend (states, validation, responsive, a11y, polish)223> - **Backend dominates ONLY:** multi-aggregate invariants, cross-service contracts, schema migrations, heavy query/perf, new event flows224>225> **Reuse-vs-Create axis (PRIMARY lever, per layer):**226>227> | UI tier | Cost |228> | -------------------------------------------- | -------- |229> | Reuse component on existing screen | 0.1-0.3d |230> | Add control/column to existing screen | 0.3-0.8d |231> | Compose components into NEW screen | 1-2d |232> | NEW screen, custom layout/states/validation | 2-4d |233> | NEW shared/common component (themed, tested) | 3-6d+ |234>235> | Backend tier | Cost |236> | ---------------------------------------------------- | --------- |237> | Reuse query/handler from new place | 0.1-0.3d |238> | Small update existing handler/entity | 0.3-0.8d |239> | NEW query on existing repo/model | 0.5-1d |240> | NEW command/handler on existing aggregate (additive) | 1-2d |241> | NEW aggregate/entity (repo, validation, events) | 2-4d |242> | NEW cross-service contract OR schema migration | 2-4d each |243> | Multi-aggregate invariant / heavy domain rule | 3-5d |244>245> **Rule:** Sum tiers across UI+backend+tests, apply productivity factor. Reuse short-circuits tiers — call out.246>247> **Test-Scope drivers (compute test_count EXPLICITLY — "+tests" hand-wave is #1 failure):**248>249> | Driver | Count |250> | --------------------------------- | ------------------------------------------------------ |251> | Happy-path journeys | 1 per story / AC main flow |252> | State-machine transitions | reachable transitions × allowed actors |253> | Multi-entity state combos | state(A) × state(B) — REACHABLE only, not Cartesian |254> | Authorization matrix | (owner, non-owner, elevated, unauth) × each mutation |255> | Validation rules | 1 per required field / boundary / format / cross-field |256> | UI states (per new screen/dialog) | happy, loading, empty, error, partial — present only |257> | Negative paths / invariants | 1 per violatable business rule |258>259> | Test tier (Trad, incl. setup+assert+flake) | Cost |260> | ------------------------------------------ | -------- |261> | 1-5 cases, fixtures reused | 0.3-0.5d |262> | 6-12 cases, 1 new fixture | 0.5-1d |263> | 13-25 cases, multi-entity setup | 1-2d |264> | 26-50 cases OR new state-machine coverage | 2-3d |265> | >50 cases OR full E2E journey | 3-5d |266>267> **Test multipliers:** new fixture/seed harness +0.5d · cross-service/bus assertion +0.3d each · UI E2E ×1.5 · each new role +1-2 cases268>269> **Blast Radius (mandatory pre-pass — affects code AND test):**270>271> 1. Files/components directly modified — count272> 2. Of those, "complex" (>500 LOC, multi-handler, central, frequently-modified) — count273> 3. Downstream consumers (callers, event subscribers, cross-service) — list274> 4. Shared/common code touched (multi-app blast) — yes/no275> 5. Regression scope — areas needing re-test276>277> **Rule:** Complex touch → add `risk_factors`. Each downstream consumer → +1-3 regression cases. Blast >5 areas OR >2 complex → re-evaluate SPLIT before estimating.278>279> **Risk Margin (drives max bound):**280>281> | likely_days | Base margin |282> | ------------------- | ------------------------------- |283> | <1d trivial | +10% |284> | 1-2d small additive | +20% |285> | 3-4d real feature | +35% |286> | 5-7d large | +50% |287> | 8-10d very large | +75% |288> | >10d | +100% AND **flag SHOULD SPLIT** |289>290> **Risk-factor add-ons (additive — enumerate in `risk_factors`):**291>292> | Factor | +margin |293> | --------------------------------------------------------------------- | ------- |294> | `touches-complex-existing-feature` (>500 LOC, multi-handler, central) | +20% |295> | `cross-service-contract` change | +25% |296> | `schema-migration-on-populated-data` | +25% |297> | `new-tech-or-unfamiliar-pattern` | +30% |298> | `regression-fan-out` (≥3 downstream areas re-test) | +20% |299> | `performance-or-latency-critical` | +20% |300> | `concurrency-race-event-ordering` | +25% |301> | `shared-common-code` (multi-consumer/multi-app) | +25% |302> | `unclear-requirements-or-design` | +30% |303>304> **Collapse rule:** total margin >100% → STOP, split (padding past 2x is dishonesty). Margin <15% on `likely_days ≥5` → under-estimated, widen.305>306> **Work-Type Caps (hard ceilings on `likely_days`):**307> | Work type | Max SP | Max likely |308> | --- | --- | --- |309> | Single field / config flag / style fix | 1 | 0.5d |310> | Add property to existing model + bind to existing UI | 2 | 1d |311> | **Additive endpoint + minor UI control** (button/menu/column), reuses fixtures | **3** | **2-3d** |312> | Additive endpoint + **NEW UI surface** OR additive multi-layer + new domain rule + 2+ test files | 5 | 3-5d |313> | NEW model/aggregate OR migration OR cross-module contract OR heavy test (>1.5d) OR NEW UI + non-trivial backend | 8 | 5-7d |314> | NEW UI surface + (NEW aggregate OR migration OR cross-service contract) | 13 | SHOULD split |315> | Cross-service contract + migration combined | 13 | SHOULD split |316> | Beyond | 21 | MUST split |317>318> **SP→Days (validation only):** 1=0.5d/0.25d · 2=1d/0.35d · 3=2d/0.65d · 5=4d/1.0d · 8=6d/1.5d · 13=10d/2.0d (Trad/AI likely)319> **AI speedup:** SP 1≈2x · 2-3≈3x · 5-8≈4x · 13+≈5x. AI cost = `(code_gen × 1.3) + (test_gen × 1.3)` (30% review overhead).320>321> **MANDATORY frontmatter:**322>323> ```yaml324> story_points: <n>325> complexity: low | medium | high | critical326> man_days_traditional: '<min>-<max>d' # range when likely ≥3d; '<N>d' when <3d327> man_days_ai: '<min>-<max>d'328> risk_margin_pct: <n> # base + add-ons329> risk_factors: [touches-complex-existing-feature, regression-fan-out] # closed-list from add-ons; [] if none330> blast_radius:331> touched_areas: <n>332> complex_touched: <n>333> downstream_consumers: [list or count]334> shared_common_code: yes | no335> estimate_scope_included: [code, integration-tests, frontend, i18n, docs]336> estimate_scope_excluded: [unit-tests, e2e, perf, deployment, code-review-rounds]337> estimate_reasoning: |338> 5-7 lines covering:339> (a) UI tier — row applied340> (b) Backend tier — row applied341> (c) Test scope — case breakdown by driver, file count, fixtures, tier row342> (d) Cost driver — dominant tier + why343> (e) Blast radius — touched, complex, regression scope344> (f) Risk factors — list driving margin; why not larger/smaller345> Example: "UI: compose Form/Table/Dialog → NEW screen (~1.5d). Backend: NEW command on existing aggregate,346> reuses validation+repo (~1d). Tests: 4 transitions × 2 actors + 3 validation + 2 UI states = 13 cases,347> 1 new fixture → tier 13-25 ~1.5d. Driver: UI composition + new states. Blast: 4 areas, 1 complex.348> Risk: base 35% + touches-complex +20% = 55% → max 3.9d → range 2.5-4d."349> ```350>351> **Sanity self-check:**352>353> - `likely_days ≥3d` and single-point? → reject, must be range354> - Margin <15% on `likely_days ≥5d`? → under-estimated, widen355> - Margin >100%? → STOP, split instead of buffer356> - Complex existing feature touched, no regression budget in `(c)`? → reject357> - Blast `>5` areas OR `>2` complex, no split discussion? → reject358> - Purely additive on existing model AND existing UI? → cap SP 3 unless tests >1.5d359> - NEW UI surface (page/complex form/dashboard)? → SP 5+ even if backend one endpoint360> - Backend cross-service / migration / multi-aggregate? → SP 8+ regardless of UI361> - `bottom_up_hours / 6` vs SP-Days disagreement >50%? → trust bottom-up, downgrade SP362> - Without tests, SP drops ≥1 bucket? → tests dominate; state explicitly363> - Reasoning called out UI vs backend vs blast vs risk factors? → if missing, add364365<!-- /SYNC:estimation-framework -->366367<!-- SYNC:ai-mistake-prevention -->368369> **AI Mistake Prevention** — Failure modes to avoid on every task:370>371> **Re-read files after context changes.** Context compaction, resume, or long-running work can make memory stale; verify current files before acting.372> **Verify generated content against source evidence.** AI hallucinates APIs, names, claims, and document facts. Check the relevant source before documenting or referencing.373> **Check downstream references before deleting or renaming.** Removing an artifact can stale docs, generated mirrors, configs, and callers; map references first.374> **Trace the full impact chain after edits.** Changing a definition can miss derived outputs and consumers. Follow the affected chain before declaring done.375> **Verify ALL affected outputs, not just the first.** One green check is not all green checks; validate every output surface the change can affect.376> **Assume existing values are intentional — ask WHY before changing OR flagging one as a defect.** Before changing or reporting a constant, limit, flag, cutoff, wording, or pattern, read nearby context and history, the CALLER's ordering, and 2+ sibling call sites of the same convention. A doc stating WHAT without WHY is missing rationale, not proof of a missing guard.377> **Surface ambiguity before acting — don't pick silently.** Multiple valid interpretations require an explicit question or stated assumption with risk.378> **Assert the outcome your system owns, not the intermediate state your infrastructure owns.** When verifying async work, assert the final business state — never the delivery/retry bookkeeping held in shared infrastructure that any co-running process can write. Such a check passes when run alone and flakes the moment anything else shares that infrastructure.379> **Keep shared guidance role-relevant.** Universal guidance must help every receiving skill or agent; code-specific obligations belong only in code-specific protocols.380381<!-- /SYNC:ai-mistake-prevention -->382383<!-- SYNC:project-protocol-overlay -->384385> **Project Protocol Overlay** — Before executing this skill, resolve any PROJECT overlay rules layered onto it: match this skill's name against the `Target` column of the project's skill-protocol index (`docs/project-reference/skill-protocols-reference.md` by default; a `referenceDocs` entry in `docs/project-config.json` overrides the path), taking the most specific matching tier ONLY — exact name > glob > `*`. **That precedence orders overlays against EACH OTHER, never against this skill.** Read ONLY the matched bodies, resolved as `<protocols-dir>/<Name>.md`; a row's Body link is display text, never a read path. A matched body that is missing or malformed is REPORTED and skipped — never reconstructed from the index Description. No index, or no match -> proceed with no overlay, silently. Full contract: `.claude/skills/project-skill-protocol/references/registry.md`.386>387> Overlays are **ADDITIVE ONLY**: they ADD rules on top of this skill's own protocol and NEVER replace, override, disable, or reinterpret a rule it already states — removing every overlay must return this skill to exactly its documented behavior. An overlay is a BRIEF, not an authority escalation: it can NEVER waive a workflow gate, git discipline, a review gate, or a user-confirmation gate. A genuine overlay-vs-skill conflict, or two equally-specific overlays that directly contradict -> surface both to the user; NEVER resolve silently.388389<!-- /SYNC:project-protocol-overlay -->390391<!-- SYNC:project-protocol-overlay:reminder -->392393**MUST ATTENTION** resolve project protocol overlays for this skill BEFORE executing — most specific matching tier only (exact > glob > `*`, which ranks overlays against each other, NEVER against this skill), read only matched bodies at `<protocols-dir>/<Name>.md`; a missing or malformed body is reported, never reconstructed. Overlays are ADDITIVE ONLY (they never replace this skill's own rules) and are a brief, NEVER an authority escalation; an equal-specificity contradiction goes to the user.394395<!-- /SYNC:project-protocol-overlay:reminder -->396397## Closing Reminders398399**Protocols in force (concise digest of the SYNC/shared blocks this skill carries):**400401- **Estimation Framework:** Bottom-up first, blast-radius pass, min-max range, tier + risk-margin tables.402- **AI Mistake Prevention:** verify generated content against evidence, trace downstream references, verify all affected outputs, re-read after context loss, surface ambiguity.403- **Critical Thinking:** Traced `file:line` proof per claim, confidence >80% to act, no guess-as-fact.404405**IMPORTANT MUST ATTENTION** compute TRUE estimate using the SAME canonical framework — fair comparison requires identical methodology406**IMPORTANT MUST ATTENTION** separate developer execution signal from model calibration signal — never collapse to single verdict407**IMPORTANT MUST ATTENTION** never claim model adjustment from a single sample — explicitly state "needs ≥3 samples for signal"408**IMPORTANT MUST ATTENTION** never trust git timestamps as coding time — always ask user to confirm/override409**IMPORTANT MUST ATTENTION** list per-layer deltas (UI/backend/tests/blast) — aggregate variance hides where model went wrong410**IMPORTANT MUST ATTENTION** use min-max ranges for both pre-impl and TRUE — comparing single points is dishonest about uncertainty411**IMPORTANT MUST ATTENTION** apply Blast Radius pass on observed diff before applying tier tables412**IMPORTANT MUST ATTENTION** persist samples to `plans/_estimation-samples.csv` for longitudinal calibration413**IMPORTANT MUST ATTENTION** state confidence per verdict — uncertainty about actual time goes in caveats414415> **[IMPORTANT]** Use `TaskCreate` to break ALL work into small tasks BEFORE starting.416417<!-- SYNC:critical-thinking-mindset -->418419> **Critical Thinking Mindset** — Apply critical thinking, sequential thinking. Every claim needs traced proof, confidence >80% to act.420> **Anti-hallucination:** Never present guess as fact — cite sources for every claim, admit uncertainty freely, self-check output for errors, cross-reference independently, stay skeptical of own confidence — certainty without evidence root of all hallucination.421422<!-- /SYNC:critical-thinking-mindset -->423424<!-- SYNC:critical-thinking-mindset:reminder -->425426**MUST ATTENTION** apply critical + sequential thinking — every claim needs appropriate traced evidence (`file:line` for repo/code claims; source URL or artifact section for research, product, content, and docs claims); confidence >80% to act, <60% DO NOT recommend. Anti-hallucination: never present guess as fact, admit uncertainty freely, cross-reference independently, stay skeptical of own confidence.427428<!-- /SYNC:critical-thinking-mindset:reminder -->429<!-- SYNC:ai-mistake-prevention:reminder -->430431**MUST ATTENTION** apply AI mistake prevention — verify generated content against evidence, trace downstream references before deleting or renaming, verify all affected outputs, re-read files after context loss, and surface ambiguity before acting.432433<!-- /SYNC:ai-mistake-prevention:reminder -->434435**[TASK-PLANNING]** Before acting, analyze task scope and systematically break it into small todo tasks and sub-tasks using TaskCreate.
duc01226