Spec Discovery

[BLOCKING] Execute skill steps in declared order. NEVER skip, reorder, or merge steps without explicit user approval. [BLOCKING] Before each step or sub-skill call, update task tracking: set in_progress when step starts, set completed when step ends. [BLOCKING] Every completed/skipped step MUST include brief evidence or explicit skip reason. [BLOCKING] If Task tools are unavailable, create and maintain an equivalent step-by-step plan tracker with the same status transitions.

Quick Summary

Goal: Before a single line of a new Feature Spec is authored, deliver the pre-spec landscape — every existing Feature Spec the idea relates to / overlaps / depends on / would affect, the related code logic, the missing features and missing test cases / user stories, the system unknowns, and the invariant landscape the new spec must respect — so the author never ships a duplicate, contradicts a [HARD] rule, or specs into a blind spot.

Summary:

• This is BOTH spec-aware and code-aware: it reads docs/specs/** (the canonical Feature Specs) AND delegates to /investigate + code-graph for the code logic the idea touches. Spec-only or code-only discovery misses half the landscape.
• It runs BEFORE spec [mode=draft] and feeds it. Its job is to decide WHETHER a new standalone spec is even the right move — the alternative is extending an existing spec, which only a spec-corpus scan can reveal.
• It is INLINE on the main agent (NOT a sub-agent) because step 5 is a BLOCKING AskUserQuestion scope-decision gate that only works inline. It MAY spawn sub-agents for parallel spec reads, but it orchestrates and gates inline.
• Greenfield short-circuit: when there are no specs AND no code, auto-detect it, record the reason, skip the heavy discovery, and hand off a minimal landscape — never grind through empty discovery.
• Main steps (0→6) — do ALL in order: (0) frame scope = keywords/entities/bucket → (1) spec-corpus discovery = Glob all candidate specs, read §1/§4/§5/§8, classify each EXTENDS/OVERLAPS/DEPENDS-ON/AFFECTED/UNRELATED with file:line → (2) code-logic discovery = /investigate + MANDATORY graph expansion, bridge code→spec via §8 [Source:] → (3) gap & invariant analysis = missing features, missing TCs/user stories, system unknowns, [HARD]/§5 invariant landscape → (4) report incrementally to plans/.../spec-discovery-{slug}.md → (5) BLOCKING AskUserQuestion scope gate = recommend NEW / EXTEND X / SPLIT, confirm cross-refs → (6) handoff to domain-analysis + spec [mode=draft|update].

Workflow:

0. Scope — read the framed capability (brainstorm/idea output); extract keywords, candidate entities/actors, target spec bucket.
1. Spec-corpus discovery — Glob docs/specs/**/README.*.md; read §1/§4/§5/§8 of each candidate; classify each as EXTENDS / OVERLAPS / DEPENDS-ON / AFFECTED / UNRELATED.
2. Code-logic discovery — /investigate {keywords} + MANDATORY graph expansion on key files when .code-graph/graph.db exists; bridge code→spec via §8 [Source:] anchors.
3. Gap & invariant analysis — missing features, missing test cases / user stories, system unknowns (<80% confidence), and the existing [HARD] rules / §5 invariants the idea must respect.
4. Report — write plans/{plan-dir}/research/spec-discovery-{slug}.md incrementally (Related Specs · Related Code · Affected Specs · Gaps · Invariant Landscape · Open Questions).
5. Scope-decision gate (BLOCKING AskUserQuestion) — recommend NEW / EXTEND existing X / SPLIT into N, and confirm which existing specs to cross-reference.
6. Handoff — feed entities, invariants, cross-refs, and gaps into domain-analysis + spec [mode=draft].

Key Rules:

• Landscape over implementation — surface related/overlapping/affected specs + the invariant landscape fast; this is NOT the spec author and NOT a deep investigation.
• INLINE execution — the step 5 user gate is BLOCKING and only works inline; spawn sub-agents only for parallel spec reads, never delegate the whole skill.
• NEVER auto-pick NEW — step 5 is a BLOCKING user gate. OVERLAPS is exactly what the spec scan exists to catch; recommend, then let the user decide scope.
• NEVER skip graph expansion when .code-graph/graph.db exists; when absent, grep + read still bridge code→spec via [Source:] anchors.

Spec-Discovery — Pre-Spec Landscape Investigation

When to Use

• About to author a NEW Feature Spec from an idea / requirement / brainstorm output, before spec [mode=draft] runs.
• Need to know whether the idea is genuinely new or overlaps an existing spec (duplicate-spec prevention).
• Need the invariant landscape — the existing [HARD] rules and §5 invariants a new capability must respect or might violate.

NOT for: authoring the spec (use spec [mode=draft]), deep root-cause analysis of existing code (use investigate), generating Section 8 test cases (use spec [mode=tests]), regenerating the derived bucket index/ERD (use spec-index).

Phase 0: Classify Corpus & Short-Circuit

Before any discovery, classify what landscape exists. This decides which steps run.

Greenfield / empty-corpus short-circuit. When Phase 0 detects no specs AND no code: record Corpus: greenfield — no specs, no code for {keywords} with the Glob/grep evidence that proved it, skip steps 1–3, write a minimal landscape report (just the framed scope + open questions), and hand off to spec [mode=draft]. Run the step 5 scope gate ONLY if there is something to decide (e.g. two plausible buckets); with nothing to decide, default to NEW and state the assumption in one line.

Workflow

Step 0: Frame the Scope

Read the framed capability — the brainstorm / idea / requirement text that triggered this. Extract:

• Keywords — domain nouns and verbs the idea names (entities, actions, features).
• Candidate entities / actors — the business objects and roles the idea implies.
• Target spec bucket — which docs/specs/{Bucket}/ the new spec would most likely live in (per the project's module mapping; resolve from docs/project-reference/feature-spec-reference.md / spec-system-reference.md).

State the framed scope in one line before continuing (e.g. Discovering for: "bulk order export" — keywords [order, export, batch], bucket Orders).

Step 0.5: Declare the Discovery Wave

Step 1 (specs), Step 2 (code), and the invariant/test-case sweep read DIFFERENT inputs and produce DIFFERENT report sections — they are PAR. Running them one after another triples the wall time of the gate that stands between an idea and a duplicate spec, for zero safety gain. Declare the wave before Step 1, then spawn its members in ONE message:

Rules binding this wave: each member owns a unique artifact path under plans/reports/spec-discovery-{slug}/; no worker writes spec-discovery-{slug}.md. After the barrier, YOU are the sole reducer: read and validate every artifact, then synthesize the final report in section order. A missing artifact is rerun or reported, never silently replaced by a bounded summary.

SEQ — keep these OUT of the wave (each names its blocker): the Step 2 graph expansion (YOU run it, and only after the code member returns its key files) · Step 3 gap & invariant reconciliation (consumes all three members) · the Step 5 scope-decision gate (a BLOCKING AskUserQuestion cannot block from inside a sub-agent) · Step 6 handoff.

Phase 0's corpus state shrinks the wave: Specs only → drop the code member · Code only → drop the corpus sweep · Greenfield → no wave at all (short-circuit).

Step 1: Spec-Corpus Discovery

# Enumerate every canonical Feature Spec
ls docs/specs/**/README.*.md 2>/dev/null   # or: Glob docs/specs/**/README.*.md

If NONE → record No existing specs and skip to Step 2.

Else, for each candidate spec the keywords touch, read the high-signal sections only (do NOT read whole specs — landscape, not deep-dive):

• §1 Overview — what the spec covers (scope boundary).
• §4 Business-Rule headers — the BR-{FC}-NN IDs and their [HARD]/[SOFT] tags (feeds invariant landscape).
• §5 Domain Model — entities + ERD (overlap detection by shared entities).
• §8 Test-Case summary — the TC count + summary table (coverage baseline; missing-TC detection).

Use the bucket INDEX.md (produced by /spec-index) as a fast navigation map when present — it lists the specs and their entities so you read fewer full files.

Classify each candidate spec's relationship to the idea (one label per spec, with file:line evidence):

Step 2: Code-Logic Discovery (only if code exists)

Bridge the idea to the implementation so the spec reflects what actually exists (or what the idea will touch).

1. Delegate to /investigate {keywords} — fast parallel file discovery of the code the idea relates to. Use investigate's numbered, prioritized list as targets; do NOT re-grep what investigate already mapped.
2. MANDATORY graph expansion — when .code-graph/graph.db exists, run graph commands YOURSELF (sub-agents cannot) on 2–3 key files investigate surfaced:

   python .claude/scripts/code_graph trace <key-entity-or-command> --direction both --json
   python .claude/scripts/code_graph connections <key-file> --json
   

   Graph reveals callers, consumers, event chains, and tests grep cannot find — exactly the downstream the new spec must account for.
3. Delegate ambiguous areas to /investigate — when investigate + graph surface a flow whose behavior is unclear (the idea hinges on how it works), hand that narrow slice to investigate rather than guessing.
4. Bridge code → spec — for each key code file, find its governing spec via the §8 [Source: namespace/service/id] anchors / Related Files. A code area with NO governing spec is a gap (record in Step 3); a code area WITH a governing spec strengthens the Step 1 relationship classification.

Step 3: Gap & Invariant Analysis

From Steps 1–2, synthesize four lists (every item file:line-cited or marked "inferred"):

• Missing features — behavior the idea implies that NO existing spec or code covers. These are the net-new surface the spec must define.
• Missing test cases / user stories — in the specs the idea touches (EXTENDS/AFFECTED), the AC / TC the idea's behavior would require but that are absent today.
• System unknowns — anything the discovery could not resolve to >80% confidence (unverified flows, ambiguous ownership, unread cross-service consumers). Name each explicitly — an unknown surfaced is cheaper than a wrong spec.
• Invariant landscape — the existing [HARD] business rules (§4) and §5 entity invariants the idea must respect or might violate. This is the single most load-bearing output: a new spec that contradicts a [HARD] rule of a DEPENDS-ON spec ships a defect. List each invariant as "for ALL {inputs}, {invariant} holds — owned by {spec/BR-id}".

Step 4: Report

Write plans/{plan-dir}/research/spec-discovery-{slug}.md (resolve {plan-dir} from the active plan; fall back to plans/reports/spec-discovery-{YYMMDD}-{HHmm}-{slug}.md when no plan dir exists). Persist incrementally — append each section as it is produced, never hold the whole report in memory:

# Spec-Discovery: {idea}

## Framed Scope
{keywords, candidate entities/actors, target bucket}

## Related Specs
| Spec | Relationship | Overlap evidence | Action implied |
| ---- | ------------ | ---------------- | -------------- |

## Related Code
{investigate's prioritized files + graph evidence — callers/consumers/tests}

## Affected Specs (forward-impact)
{specs whose documented behavior the idea would change}

## Gaps
- Missing features: ...
- Missing TCs / user stories: ...

## Invariant Landscape
- for ALL {inputs}, {invariant} — owned by {spec/BR-id} — idea must {respect/extend}

## Open Questions
- {system unknowns, <80% confidence items}

Step 5: Scope-Decision Gate (BLOCKING AskUserQuestion)

MANDATORY MUST ATTENTION — NO EXCEPTIONS: before any spec is authored, MUST ATTENTION use AskUserQuestion to present the recommended scope. NEVER auto-pick — OVERLAPS detection is the whole reason this skill exists; assuming NEW silently ships duplicates.

Recommend ONE option (with the evidence behind it) and confirm the cross-references:

• (a) NEW standalone spec — no EXTENDS/OVERLAPS match; the idea is genuinely net-new. Hand off to spec [mode=draft].
• (b) EXTEND existing spec X — an EXTENDS/OVERLAPS match means the idea belongs inside X. Reroute to /spec [mode=update] against X instead of drafting a new file.
• (c) SPLIT into N specs — the idea spans N distinct capabilities (or would breach the size caps); author N specs, each with its own bucket.

Also confirm WHICH existing specs (the DEPENDS-ON / AFFECTED set) the author must cross-reference, so the new/updated spec links them and respects their invariants.

Step 6: Handoff

Feed the discovery forward:

• → domain-analysis — the related entities + invariant landscape (so the domain model is consistent with existing specs).
• → spec [mode=draft] (or spec [mode=update] if Step 5 chose EXTEND) — the framed scope, the missing features/TCs, the cross-references to link, and the [HARD] rules to respect.

Results Format

## Spec-Discovery Results: {idea}

### Recommended Scope
**{NEW | EXTEND spec X | SPLIT into N}** — because {evidence-backed reason}

### Related Specs
| Spec | Relationship | Evidence | Implied action |
| --- | --- | --- | --- |
| `docs/specs/{Bucket}/README.{X}.md` | OVERLAPS | §4 BR-X-03 already states this | route to /spec mode=update |

### Related Code
1. `{file}` — {role} — graph: {callers/consumers found}

### Affected Specs (forward-impact)
- `{spec}` — {documented behavior the idea changes}

### Gaps
- Missing features: {list}
- Missing TCs / user stories: {list}

### Invariant Landscape (must respect)
- for ALL {inputs}, {invariant} — owned by {spec/BR-id}

### Open Questions
- {system unknowns, <80% confidence}

**Full report:** plans/{plan-dir}/research/spec-discovery-{slug}.md

Related Skills

investigate (code discovery and deep-dive ambiguous flows — Step 2) | spec [mode=draft] (authors §1–7 from the idea — Step 6 handoff) | spec [mode=update] (the EXTEND reroute — Step 5) | domain-analysis (entity/invariant modeling — Step 6) | spec-index (derived bucket INDEX.md used in Step 1)

[IMPORTANT] Use TaskCreate to break ALL work into small tasks BEFORE starting — including a task per candidate spec read. This prevents context loss from long specs. For trivial single-spec scopes, AI MUST ATTENTION ask user whether to skip.

• docs/project-reference/feature-spec-reference.md — Feature Spec conventions, bucket/module mapping (read before reading any spec).
• docs/project-reference/spec-system-reference.md — canonical vs derived spec artifacts, TC format, spec paths.
• docs/project-reference/domain-entities-reference.md — Domain entity catalog, relationships, cross-service sync (read when the idea involves business entities/models).

External Memory: Complex/lengthy discovery → write findings incrementally to the research report. Prevents context loss.

Evidence Gate: MANDATORY MUST ATTENTION — every relationship classification, gap, and invariant requires file:line proof with confidence % (>80% act, <80% verify first).

Graph-Assisted Investigation — MANDATORY when .code-graph/graph.db exists.

HARD-GATE: MUST ATTENTION run at least ONE graph command on key files before concluding any investigation.

Pattern: Grep finds files → trace --direction both reveals full system flow → Grep verifies details

Task	Minimum Graph Action
Investigation	trace --direction both on 2-3 entry files
Fix/Debug	callers_of on buggy function + tests_for
Feature/Enhancement	connections on files to be modified
Code Review	tests_for on changed functions
Blast Radius	trace --direction downstream

CLI: python .claude/scripts/code_graph {command} --json. Use --node-mode file first (10-30x less noise), then --node-mode function for detail.

Incremental Result Persistence — MANDATORY for all sub-agents or heavy inline steps processing >3 files.

1. Before starting: Create report file plans/reports/{skill}-{date}-{slug}.md and record Run ID, Task ID, Attempt ID, target scope, and target fingerprint.
2. After each file/section reviewed: Append findings, evidence, changed paths, and gaps immediately — never hold them in memory.
3. Delegated return: A sub-agent emits only the structured SYNC:subagent-return-contract envelope with exact totals, salient Critical/High findings (maximum ten), current attempt, and Full report: path. Inline user-facing output: Preserve the skill's requested explanation or teaching, with links to the persisted evidence; the delegated transport limit does not replace that deliverable. Do not paste a full review report into an envelope.
4. Parent synthesis: The main agent reads the full report for synthesis, acceptance, deduplication, and repair planning — not only when a named blocker exists. It preserves all severities beyond the transport cap.
5. Read-only boundary: A read-only leaf may write its report/repair proposal but MUST NOT edit source, generated output, or user data; the parent/owner performs repairs after acceptance.
6. Advancement gate: The parent records ACCEPTED for the current Attempt ID only after reconciling target, totals, gaps, and changed paths; stale or late attempts cannot advance dependent work.

Why: Context cutoff mid-execution loses ALL in-memory findings. Each disk write survives compaction. Partial results are better than no results, while explicit identity prevents a late result from being mistaken for the current run.

Report naming: plans/reports/{skill-name}-{YYMMDD}-{HHmm}-{slug}.md

Sub-Agent Return Contract — When this skill spawns a sub-agent, the sub-agent MUST return ONLY the structured envelope below. Main agent reads the envelope first, then opens the referenced report for synthesis, acceptance, deduplication, or repair planning; a full report is never pasted inline.

## Sub-Agent Result: [skill-name]

Status: ✅ PASS | ⚠️ PARTIAL | ❌ FAIL
Confidence: [0-100]%
Run ID: [stable run identifier]
Task ID: [parent task or phase identifier]
Attempt ID: [monotonic attempt/revision identifier]
Target: [exact files/paths or scope] @ [target fingerprint/commit]
Changed paths: [none | exact paths]
Finding totals: Critical=[n] | High=[n] | Medium=[n] | Low=[n]
Acceptance: PENDING | ACCEPTED | REJECTED — parent records the decision

### Findings (Critical/High surfaced — max 10 bullets)

- [severity] [file:line] [finding]

### Gaps / Unverified

- [missing host, runtime, coverage, or evidence limitation]

### Actions Taken

- [file changed] [what changed]

### Blockers (if any)

- [blocker description, or `none`]

Full report: plans/reports/[skill-name]-[date]-[slug].md

The ten-bullet limit is a transport limit, not a visibility limit: the full report may contain more than ten Medium/Low findings when no named blocker exists, and the parent MUST read it when synthesizing or deduplicating. The parent MUST reject a stale, duplicate, or superseded Attempt ID and MUST accept the current attempt before advancing a dependent step. Read-only leaves write repair proposals/reports only; they do not edit source, generated carriers, or user files.

Context budget — the return payload is a SUMMARY, not a transcript: no raw file contents / full diffs / verbatim logs inline, no re-pasted source. Everything beyond the envelope lives in the incrementally-written report. A sub-agent that would exceed the summary shape MUST persist the detail and return only the pointer; bounded transport must never become bounded visibility.

Nested Task Expansion Contract — For workflow-step invocation, the [Workflow] ... row is only a parent container; the child skill still creates visible phase tasks.

1. Call TaskList first. If a matching active parent workflow row exists, set nested=true and record parentTaskId; otherwise run standalone.
2. Create one task per declared phase before phase work. When nested, prefix subjects [N.M] /skill-name — phase.
3. When nested, link the parent with TaskUpdate(parentTaskId, addBlockedBy: [childIds]).
4. Orchestrators must pre-expand a child skill's phase list and link the workflow row before invoking that child skill or sub-agent.
5. Mark exactly one child in_progress before work and completed immediately after evidence is written.
6. Complete the parent only after all child tasks are completed or explicitly cancelled with reason.

Blocked until: TaskList done, child phases created, parent linked when nested, first child marked in_progress.

Project Reference Docs Gate (static JIT) — Run after task-tracking bootstrap and immediately before target/source file reads, grep, edits, tests, or analysis. Project docs override generic framework assumptions; hooks may remind or accelerate this gate, but never prove that it ran.

1. Identify scope: file types, domain area, and operation.
2. Read docs/project-config.json first — the project's machine-readable map. It is the single source of truth for THIS repo (modules/paths, framework + search keywords, test/E2E/integration run-commands, design system, architecture rules, workflow patterns); ground exact paths, run-commands, and conventions on it before investigating, planning, or coding — never assume framework defaults (CLAUDE.md + reference docs are derived from it). If it — or the docs index, lessons.md, CLAUDE.md, AGENTS.md, or any required reference doc — is missing or stale, auto-run /project-init or the narrow route (/project-config, /docs-init, /scan-all, /scan --target=<key>, /claude-md-init) first; if Codex mirrors or AGENTS.md are stale, ask the user to run /sync-codex (never auto-run it).
3. Required docs by trigger: always docs/project-reference/lessons.md; doc lookup docs-index-reference.md; review code-review-rules.md; backend/CQRS/API backend-patterns-reference.md; domain/entity domain-entities-reference.md; frontend/UI frontend-patterns-reference.md; styles/design scss-styling-guide.md + design-system/design-system-canonical.md; integration tests integration-test-reference.md; E2E e2e-test-reference.md; feature docs/specs feature-spec-reference.md + spec-system-reference.md + spec-principles.md; behavior/public-contract/spec-test-code sync workflow-spec-test-code-cycle-reference.md; derived spec index/ERD/reimplementation guides spec-system-reference.md + source Feature Specs under docs/specs/; architecture/new area project-structure-reference.md.
4. Read every required doc, then before target work state: Reference docs read: ... | Not applicable: .... After compaction, resume, delegation, or a material context change, repeat the route and restate the set; prior conversation and hook output are not proof of current loading.

Ready when: scope evaluated, docs/project-config.json consulted, required docs checked/read or setup route completed, lessons.md confirmed, citation emitted.

Task Tracking & External Report Persistence — Bootstrap this before execution; then run project-reference doc prefetch before target/source work.

1. Create a small task breakdown before target file reads, grep, edits, or analysis. On context loss, inspect the current task list first.
2. Mark one task in_progress before work and completed immediately after evidence; never batch transitions.
3. For plan/review work, create plans/reports/{skill}-{YYMMDD}-{HHmm}-{slug}.md before first finding.
4. Append findings after each file/section/decision and synthesize from the report file at the end.
5. Final output cites Full report: plans/reports/{filename}.

Blocked until: task breakdown exists, report path declared for plan/review work, first finding persisted before the next finding.

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.

Evidence-Based Reasoning — Speculation is FORBIDDEN. Every claim needs proof.

1. Cite file:line, grep results, or framework docs for EVERY claim
2. Declare confidence: >80% act freely, 60-80% verify first, <60% DO NOT recommend
3. Cross-service validation required for architectural changes
4. "I don't have enough evidence" is valid and expected output

BLOCKED until: - [ ] Evidence file path (file:line) - [ ] Grep search performed - [ ] 3+ similar patterns found - [ ] Confidence level stated

Forbidden without proof: "obviously", "I think", "should be", "probably", "this is because" If incomplete → output: "Insufficient evidence. Verified: [...]. Not verified: [...]."

Cross-Service Check — Microservices/event-driven: MANDATORY before concluding investigation, plan, spec, or feature doc. Missing downstream consumer = silent regression.

Boundary	Grep terms
Event producers	Publish, Dispatch, Send, emit, EventBus, outbox, IntegrationEvent
Event consumers	Consumer, EventHandler, Subscribe, @EventListener, inbox
Sagas/orchestration	Saga, ProcessManager, Choreography, Workflow, Orchestrator
Sync service calls	HTTP/gRPC calls to/from other services
Shared contracts	OpenAPI spec, proto, shared DTO — flag breaking changes
Data ownership	Other service reads/writes same table/collection → Shared-DB anti-pattern

Per touchpoint: owner service · message name · consumers · risk (NONE / ADDITIVE / BREAKING).

BLOCKED until: Producers scanned · Consumers scanned · Sagas checked · Contracts reviewed · Breaking-change risk flagged

Rationalization Prevention — AI skips steps via these evasions. Recognize and reject:

Evasion	Rebuttal
"Too simple for a plan"	Simple + wrong assumptions = wasted time. Plan anyway.
"I'll test after"	RED before GREEN. Write/verify test first.
"Already searched"	Show grep evidence with file:line. No proof = no search.
"Just do it"	Still need TaskCreate. Skip depth, never skip tracking.
"Just a small fix"	Small fix in wrong location cascades. Verify file:line first.
"Code is self-explanatory"	Future readers need evidence trail. Document anyway.
"Combine steps to save time"	Combined steps dilute focus. Each step has distinct purpose.

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.

MUST ATTENTION cite file:line evidence for every claim. Confidence >80% to act, <60% = do NOT recommend.

MUST ATTENTION never skip steps via evasions. Plan anyway. Test first. Show grep evidence with file:line.

MUST ATTENTION run at least ONE graph command on key files before concluding when .code-graph/graph.db exists.

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.

• MANDATORY Bootstrap task tracking before target work; transition one task at a time.
• MANDATORY Persist plan/review findings to plans/reports/ incrementally and synthesize from disk.

• MANDATORY Before investigating, planning, or coding, read docs/project-config.json (the project map: modules/paths, run-commands, conventions, architecture/workflow rules) + the required project-reference docs, and cite Reference docs read: ....
• MANDATORY Load detail just in time immediately before the first target read/grep/edit/test; hooks may provide a pointer, but a hook event or prior turn is never evidence that the current files were read.
• MANDATORY Always include lessons.md; project config + conventions override generic framework defaults.
• MANDATORY If project config, root instruction files, or any required reference doc is missing or stale, auto-run /project-init or the narrow lower-level route before ordinary project-specific work. On compaction, resume, delegation, or a context change, re-read the required docs and restate the route before continuing.

• MANDATORY Parent workflow rows do not replace child phase tracking; expand phases and link the parent when nested.
• MANDATORY Orchestrators pre-expand child skill phases before invocation; use [N.M] /skill-name — phase prefixes and one-in_progress discipline.

Prompt-Enhance Closing Anchors

IMPORTANT MUST ATTENTION follow declared step order for this skill; NEVER skip, reorder, or merge steps without explicit user approval IMPORTANT MUST ATTENTION for every step/sub-skill call: set in_progress before execution, set completed after execution IMPORTANT MUST ATTENTION every skipped step MUST include explicit reason; every completed step MUST include concise evidence IMPORTANT MUST ATTENTION if Task tools unavailable, maintain an equivalent step-by-step plan tracker with synchronized statuses

Parallel Sub-Agent Dispatch — Plan parallelism the moment a task breakdown exists, BEFORE executing it — running provably independent tasks sequentially wastes wall-clock. Applies to every multi-step job: workflow steps, planning, batch updates, investigation, research, scans, reviews, doc sync. Plan execution is metadata-gated, NEVER default-parallel — fan-out follows ONLY what the plan declares (PAR/SEQ tags + per-phase write set); an untagged plan runs sequentially — why: a derived write set cannot see cascade or generated writes.

1. Tag every task PAR or SEQ. PAR = inputs exclude every pending task's output AND write set disjoint from every other PAR. Else SEQ — MUST ATTENTION name the dependency forcing it.
2. Group PAR into waves. No edge between members. Two writers of one file NEVER share a wave. Read-only work (search, investigation, review, research) parallelizes freely.
3. Declare before dispatch: Parallel plan: wave 1 = [...] · wave 2 = [...] · SEQ = [...] (reason).
4. Spawn each wave in ONE message — every Agent call in one response, NEVER dripped per turn. Route each task to its specialist (.claude/skills/shared/sub-agent-selection-guide.md); NEVER code-reviewer as catch-all.
5. Brief each sub-agent self-contained: goal · scope + owned files · reference docs · return contract (summary + Full report: path, per SYNC:subagent-return-contract) · incremental persistence to plans/reports/ (per SYNC:incremental-persistence).
6. Barrier per wave. Advance ONLY after EVERY member returns (a skipped conditional counts as returned). Merge, mark each task completed/skipped, THEN dispatch the next wave. Mutating steps wait for the barrier.
7. One level deep. A dispatched sub-agent executes its own brief; further fan-out stays the orchestrator's job unless that agent's .claude/agents/*.md definition authorizes it.

NEVER parallelize: tasks sharing a write target · a task consuming a pending task's outpu

…(truncated)