[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_progresswhen step starts, setcompletedwhen 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: Drive a diff scope to a clean pass by pairing /changes-review with /fix in a recursive loop — each round runs /changes-review INLINE in report-only mode to surface validated findings over a fixed diff scope, then /fix to resolve them at the owning layer, then loops again with a FRESH full /changes-review over the CHANGED diff — stopping when a complete /changes-review pass clears the round's exit bar: zero validated findings in round 1, and zero validated CRITICAL/HIGH/MEDIUM from round 2 (LOW-only ENDS the loop, deferred not fixed).
Summary:
- Each round =
/changes-review(report-only) + validate +/fix— the loop runs/changes-reviewas a review-PRODUCER that stops after its raw findings report (the documented/workflow-review-changesboundary: "stop after the report; parent step 2 owns validation",changes-review/SKILL.md:52,204); the loop then owns the validation gate (/why-review --validate-findings) and the dedicated/fixhalf, so/changes-reviewnever self-validates or self-fixes; one without the other never converges. - Steps (in order): (0) resolve diff scope + Goal Contract → (0b) bind the convergence loop (protocol loop primary + optional
/goalaccelerator) → (1) round loop { run/changes-reviewreport-only INLINE →/why-review --validate-findingson the report → run/fixon the VALIDATED BLOCKING findings at the owning layer → log iteration } → (2) converge when a fresh review clears the round's exit bar (zero findings round 1; zero CRITICAL/HIGH/MEDIUM round 2+) OR escalate on non-progress → (3) terminal/docs-update+ recap. - Convergence: stop ONLY when a fresh full
/changes-reviewover the CURRENT (post-fix) diff clears the round's exit bar — not a stale clean report predating the last fix. - Severity floor — from round 2, LOW stops blocking. Round 1 converge on zero validated findings (any severity). From round 2 the bar is zero validated CRITICAL/HIGH/MEDIUM — a round whose validated findings are ALL LOW ENDS the loop. Never open another round to fix LOW alone; list every deferred LOW in the recap and Goal Contract instead, and NEVER re-tier a real CRITICAL/HIGH/MEDIUM down to LOW to reach the exit. Severity tiers per
SYNC:severity-rubric. - Inline invariant: run
/changes-reviewand/why-reviewvia theSkilltool, NEVER theAgenttool — they self-bind their OWN review-loop obligations (and a session/goalgate WHEN available,changes-review/SKILL.md:197-223,why-review/SKILL.md:57-76), which a sub-agent cannot own or carry back to this loop./changes-review's own dimensional reviewers (its Phase 0.7 sub-agents) stay sub-agents by its design, so context stays bounded. - Apply ONLY validated findings: every finding is validated to the ≥85% survival bar via
/why-review --validate-findingsbefore/fixtouches it; the loop applies THOSE at the lowest owning layer (Entity > Service > Handler), routed by change type — NEVER unvalidated findings. - Scope base is FIXED; the working tree grows. Recompute the diff scope each round (
branch-diff base∪ current uncommitted changes) so convergence is measured against a stable subject as fixes accumulate. - Bounded: round cap default 3; blocking findings not shrinking across 2 rounds, or cap hit with CRITICAL/HIGH/MEDIUM still open → STOP & escalate via
AskUserQuestion. Increasing findings → STOP (fixes regressing).
Why this skill exists (READ FIRST — it is the whole justification): standalone /changes-review COUPLES find + fix inside one skill invocation (its Phase 7 self-fix and Phase -1 self-recursive loop, changes-review/SKILL.md:53,197-223) — the review and the fix share one context, one lens, and one confirmation bias. This loop decouples them: it runs /changes-review purely as a finder (report-only, the documented /workflow-review-changes boundary where it stops after the report and the caller owns fixing, changes-review/SKILL.md:29,204,558,578), validates the findings, then hands them to a dedicated /fix half with its own intelligent routing at the lowest owning layer, and then re-runs a fresh full /changes-review over the changed diff. The value is the clean finder/fixer split plus the fresh-full re-review each round: it catches fix-induced regressions the coupled inner loop can rationalize away, and it lets /fix own the fix mechanics instead of the reviewer patching its own findings. Without this outer loop, "changes-review found issues, then something fixed them" ships those fixes without an independent fresh review.
Workflow: resolve diff scope + Goal Contract → bind the convergence loop (protocol loop + optional /goal accelerator) → round loop { run /changes-review report-only INLINE → /why-review --validate-findings → run /fix on the validated findings at owning layer → log iteration } → converge when a fresh full review clears the round's exit bar (zero validated findings round 1; zero validated CRITICAL/HIGH/MEDIUM round 2+) → terminal /docs-update + recap.
Key Rules:
- Each round pairs
/changes-review(find) +/fix(resolve). The loop runs/changes-reviewin report-only mode so it produces findings but does NOT self-validate or self-fix; the loop's/why-review --validate-findingsgate and/fixare the halves that validate and land the change. A round is complete when BOTH have run, or when the current round's bar is already clear (round 1: zero findings; round 2+: zero CRITICAL/HIGH/MEDIUM, with LOW deferred). - MUST run INLINE in the main session — NEVER dispatch
/changes-reviewor/why-reviewas a sub-agent. They self-bind their own review-loop obligations (and a session/goalgate when available,changes-review/SKILL.md:197-223); as a sub-agent that in-session guarantee is silently lost. This loop skill therefore also runs inline. (/changes-review's internal Phase 0.7 dimensional reviewers remain sub-agents by its own design — that is bounded and correct.) - Report-only invocation is mandatory. Tell
/changes-reviewto run as a review producer: do its full dimensional review + Phase 6/why-review --validate-findingsgate, then STOP before Phase 7 self-fix / Phase 7.5 holistic / Phase 8 docs-update. The loop owns fixing (Step 1.3) and the terminal docs-update (Step 3). If a/changes-reviewinvocation cannot be constrained to report-only in this environment and self-fixes anyway, fall back to detecting fixes-applied per round (like/workflow-review-changes-loop) and skip the redundant/fixhalf for that round — never double-fix. - Convergence = a fresh full
/changes-reviewover the post-fix diff clears the round's exit bar. Round 1: zero validated findings. Round 2+: zero validated CRITICAL/HIGH/MEDIUM — LOW-only converges (record the LOWs as deferred, do not fix them). A clean report produced BEFORE the latest fix landed does NOT count — re-review the changed diff. - The severity floor bounds ITERATION, never the standard. It ends the loop; it never authorizes shipping a known CRITICAL/HIGH/MEDIUM, never lowers the ≥85% finding-survival bar, and never applies to a binary gate (a failing test is a failure, not a LOW finding).
/fixapplies ONLY validated findings, at the lowest owning layer, routed by change type (code →/fixwith its--targetintelligent routing, or a direct edit at Entity/Service; spec-drift →/spec [update]+/spec [mode=tests]; docs →/docs-update; missing coverage →/integration-test; honor each finding's dual-feedback ledger,changes-review/SKILL.md:530). NEVER apply an unvalidated or demoted finding.- The diff scope base is FIXED across rounds; its content changes as fixes land. Recompute
{scope}= branch-diff base ∪ current uncommitted changes each round so the base merge-base never moves and convergence is measured against a stable subject. - Round cap (default 3) and blocking-findings-not-shrinking / increasing → STOP & escalate via
AskUserQuestion. NEVER loop open-ended. Cap exhaustion escalates only when CRITICAL/HIGH/MEDIUM remain — a LOW-only round converges via the severity floor.
First Principle — Convergence, Not Motion
A round that changes the diff is progress only if the next fresh review finds fewer things to fix. The loop exists to reach a fixed point (no blocking findings), not to keep editing the code. The bar tightens by round: everything blocks in round 1; from round 2 only CRITICAL/HIGH/MEDIUM block, so a LOW-only round is the fixed point. If findings stop shrinking, that is a signal to escalate, not to spin another round.
Step 0 — Resolve Diff Scope + Goal Contract (FIRST ACTION)
- Parse the review scope from the user prompt into a stable, reusable scope string — exactly the diff kinds
/changes-reviewresolves (changes-review/SKILL.md:96-104). It has two parts UNIONed:- Branch-diff base — a branch-to-branch or PR diff (e.g.
git diff develop...HEAD, three-dot: changes on the feature branch since it forked fromdevelop) so the base is a fixed merge-base, not a moving target. If the prompt names a commit range, capture it the same way. - Current changes — the uncommitted working-tree changes (
git status --porcelain,git diff+git diff --staged). - Scope string (recompute each round):
{branch-diff base} ∪ {current uncommitted changes}. The base commit is fixed for the whole loop; the uncommitted set legitimately grows as fixes land. - If the prompt names no branch diff (pure "current changes" review), the scope is just the working-tree changes — the loop still applies. NEVER silently convert the diff source type.
- Branch-diff base — a branch-to-branch or PR diff (e.g.
- Resolve/create the Goal Contract per
SYNC:goal-contract-satisfaction-loop(plans/goals/{YYMMDD-HHmm}-{slug}/goal.md, template.claude/templates/goal-contract-template.md). Its single required Success Criterion:A fresh full
/changes-reviewover{scope}clears the round's exit bar: round 1 → zero validated findings (no finding of any severity survives the loop's/why-review --validate-findingsgate); round 2+ → zero validated CRITICAL/HIGH/MEDIUM findings, with any remaining LOW findings recorded as deferred rather than fixed. Record the round cap (default 3) and the scope string in Constraints.
Step 0b — Bind the Convergence Loop (protocol-first; /goal is an optional accelerator)
The convergence loop is bound by TWO layers. The protocol loop (Steps 1–2) is the BINDING mechanism and MUST be self-driven by you, the running agent, on every host — with or without any command or hook. The /goal command is an OPTIONAL accelerator layered on top; it is never the primary mechanism, and its absence NEVER weakens the loop. This mirrors the project rule that hooks/trackers are accelerators only — correctness must not depend on them.
1. Protocol loop — ALWAYS binding (hook/command-independent). You are personally responsible for not stopping until the loop converges or bounded-escalates. This binds Claude, Codex, and Copilot equally, whether or not /goal exists:
Repeatedly run
/changes-reviewreport-only INLINE over{scope}(recomputed each round). After each review, validate its findings with/why-review --validate-findings, apply only the VALIDATED findings that block the current round via/fixat their owning layer, then re-run a FRESH full/changes-reviewover the CHANGED diff. Do NOT stop while the last review still produced findings that BLOCK at the current round's bar. Converge when a fresh full/changes-reviewclears that bar: round 1 → zero validated findings; round 2+ → zero validated CRITICAL/HIGH/MEDIUM (LOW-only ENDS the loop, with the LOWs recorded as deferred). Cap at{N=3}rounds; if blocking findings do not shrink across 2 consecutive rounds, findings increase, or the cap is hit with CRITICAL/HIGH/MEDIUM still open → STOP and escalate viaAskUserQuestion. Never loop open-ended.
Treat this as a standing obligation you re-read at every Step 2 checkpoint — NOT a one-time note you can rationalize away after the first fix cycle. The Goal Contract's required Success Criterion (Step 0) is its durable, host-independent record.
2. /goal command — invoke as an accelerator WHEN AVAILABLE. If a /goal command exists and you are permitted to run it in this environment, ALSO invoke it (a real tool/command call, NOT a paraphrase, NOT a Goal Contract file substituted for it) with the SAME condition, so a session Stop hook mechanically enforces the loop:
/goal changes-review convergence loop: repeatedly run /changes-review report-only INLINE over {scope} (recomputed each round). After each review, validate its findings with /why-review --validate-findings, apply only VALIDATED findings that block the current round via /fix at their owning layer, then re-run a FRESH full /changes-review over the CHANGED diff. If blocking validated findings>0 → apply fixes and run another round; if a fresh full /changes-review clears the round's bar (round 1: zero validated findings; round 2+: zero validated CRITICAL/HIGH/MEDIUM, LOW-only counts as clear with the LOWs recorded as deferred) → CONVERGED, run the terminal /docs-update and clear the gate. Do NOT stop while the last review still produced blocking validated findings. Cap at {N=3} rounds; if blocking findings do not shrink across 2 consecutive rounds, findings increase, or the cap is hit with CRITICAL/HIGH/MEDIUM still open → STOP and escalate via AskUserQuestion. Never loop open-ended.
The /goal Stop hook blocks stopping until the condition holds and auto-clears when met — do not tell the user to clear it.
If /goal is unavailable, unregistered, or not permitted (e.g. Codex/Copilot, or a Claude run without the command): DO NOT error, DO NOT block, and DO NOT invent a stand-in gate. Record ONE line in the Goal Contract — /goal accelerator unavailable — loop bound by protocol (Steps 1–2) + this Goal Contract — and proceed. The protocol loop above plus the Goal Contract are the same gate, enforced by discipline instead of a hook.
Nested gates (by design, safe): each inner
/changes-reviewround would self-bind its OWN Phase -1 review-loop obligation — but the loop runs it in report-only mode, where Phase -1 is deferred to the caller (the/workflow-review-changesboundary,changes-review/SKILL.md:204), so no inner self-fix gate is installed; THIS outer loop owns the single convergence gate. Each inner/why-review --validate-findingsself-clears when that round's findings are all adjudicated. All self-clear on satisfaction — no orphaned gate. Do NOT tell the user to clear either.
Step 1 — Round Loop (/changes-review → validate → /fix → log)
Each round couples the two halves — review to find, fix to resolve. For each round R (starting at 1), do ALL of:
- Snapshot before: record the working-tree fingerprint —
git status --porcelain+git diff --stat. This is the fixes-applied baseline for the round and the objective backstop for convergence detection (Step 2). - Run
/changes-reviewreport-only INLINE on the recomputed{scope}via theSkilltool (NEVER theAgenttool). Direct it to run as a review PRODUCER: perform its full dimensional review and STOP after writing its findings report — do NOT run its Phase 6 validation, Phase 7 self-fix, Phase 7.5 holistic, or Phase 8 docs-update. This is exactly the/workflow-review-changesboundary where the caller (this loop) owns validation and fixing, so/changes-reviewhands the raw report back and stops (changes-review/SKILL.md:52,204,558,578). If the environment cannot constrain/changes-reviewto report-only and it self-fixes anyway, treat that self-fix as this round's fix half (detect fixes-applied vs the before-snapshot) and skip Steps 1.3–1.4 to avoid double-fixing. - Validate the findings — THE LOOP OWNS THIS GATE. Run
/why-review --validate-findings <report-path>INLINE via theSkilltool over the report/changes-reviewjust produced (plans/reports/code-review-*.md), so every finding is confirmed correct, proof-backed, reasonable, and best-practice (≥85% survival bar) before any fix. This is the parent-owned validation the report-only boundary defers to the caller (changes-review/SKILL.md:52). If the validated finding set is empty → this round converged; go to Step 2 (no fix half needed). - Run
/fixon the validated blocking findings (findings>0 only) — this is the half the loop owns. Resolve each validated blocking finding at its owning layer: code →/fix(its--targetintelligent routing) or a direct edit at the lowest layer (Entity > Service > Handler); spec-drift (CODE-WRONG/SPEC-STALE/SPEC-SILENT) → route per the finding's Spec Drift Adjudication verdict (/spec [update]+/spec [mode=tests]); missing coverage (GAP/SPEC-GAP) →/integration-test; docs → defer to the terminal/docs-update; behavior-changing → honor the finding's Dual-Feedback Ledger (spec verdict + test action,changes-review/SKILL.md:530). Fix ONLY validated findings that block this round — never an unvalidated, demoted, or round-2+ LOW-only finding. - Append an Iteration Log entry to the Goal Contract: round number, validated findings count, files/artifacts changed this round (
file:line), fixes applied, and remaining gaps.
Step 2 — Convergence & Escalation Gate
Evaluate after every round:
| Condition | Action |
|---|---|
Fresh full /changes-review returned zero validated findings AND the working tree is unchanged by that final review pass |
CONVERGED → mark the required criterion PASS in the Goal Satisfaction matrix → clear the /goal gate → go to Step 3. |
Validated findings > 0 AND round < N AND findings shrank vs prior round |
Apply the validated fixes (Step 1.4), recompute {scope}, then run round R+1 (fresh full re-review of the changed diff). |
| Findings did not shrink across 2 consecutive rounds (same/increasing count) | STOP & escalate via AskUserQuestion — a non-converging loop is a signal, not a reason to spin. |
Round R ≥ 2 AND the fresh review's validated findings are ALL LOW (zero CRITICAL/HIGH/MEDIUM) |
CONVERGED on the severity floor → do NOT run another round for LOW alone → record every remaining LOW as a deferred finding in the recap + Goal Contract → mark the required criterion PASS → go to Step 3. |
Round cap N hit with CRITICAL/HIGH/MEDIUM still open |
STOP & escalate via AskUserQuestion — report the still-open findings; do not silently continue. (LOW-only at the cap converges via the severity-floor row above.) |
Increasing findings = STOP. If round
Rsurfaces MORE validated findings than roundR-1, the fixes are regressing the code — STOP and escalate immediately (mirrorschanges-review/SKILL.md's regression stance). Never trade one fix for two new findings across rounds.
Step 3 — Terminal Docs-Update + Recap
- Terminal
/docs-update(MANDATORY once converged). The loop ran/changes-reviewin report-only mode, so its unconditional Phase 8 docs-update (changes-review/SKILL.md:55,68) was deferred to the caller. Once a fresh full/changes-reviewclears the round's exit bar (zero validated findings in round 1, or zero validated CRITICAL/HIGH/MEDIUM from round 2 onward with LOW deferred), invoke/docs-updateINLINE over the full changeset as the terminal step so no stale docs survive. - Recap. Emit a concise convergence recap: rounds run, validated findings per round (the shrinking sequence), the fixes applied at each round, the final zero-findings evidence, and the Goal Satisfaction matrix (required criterion PASS). Point to each round's
/changes-reviewreport underplans/reports/and the Goal Contract Iteration Log. Do NOT commit or push unless the user explicitly asks.
Convergence Detection — Why a Fresh Full Re-Review Is Required
A round converges ONLY when a /changes-review that ran over the current, post-fix diff clears the round's exit bar — zero validated findings in round 1, zero validated CRITICAL/HIGH/MEDIUM from round 2. Both properties are required because:
- Fresh over the changed diff — a clean report from a review that predates the last fix proves nothing about the fix. Every applied fix invalidates the prior verdict (
changes-review/SKILL.md:72); the loop MUST re-review after fixing, never reuse a stale clean report. - Zero validated findings — the loop's
/why-review --validate-findingsgate already dropped inflated/unproven findings below the ≥85% bar; convergence rides on that validated set, so the loop never chases a nit the review itself would demote. - Working-tree unchanged backstop — the objective
git-diff comparison (Step 1.1 snapshot) confirms the converging review actually landed no fix; a "clean" verdict that still mutated files means the round DID fix things → run another round to re-prove clean.
When validated findings remain but cannot be fixed (owner/product input needed) → escalate, do not loop. When a fix lands but the next fresh review still finds issues → run another round. Convergence is a fixed point, not a single clean read.
IMPORTANT MANDATORY sequence: Step 0 (resolve diff scope + Goal Contract) → Step 0b (bind the convergence loop: protocol loop primary + optional /goal accelerator) → Step 1 (round loop: run /changes-review report-only INLINE → /why-review --validate-findings → run /fix on validated findings → log) → Step 2 (converge on a zero-findings fresh review / escalate on non-progress) → Step 3 (terminal /docs-update + recap).
Goal Contract Satisfaction Loop — Persist the user goal in an external file, execute against it, and loop review/fix until every saved required criterion passes or a blocker escalates. Bounded closed loop — NEVER open-ended autonomous exploration.
- Resolve the active goal (in order): active plan
goal.md→plans/goals/{YYMMDD-HHmm}-{slug}/goal.md→ create a new Goal Contract from the current user request (template:.claude/templates/goal-contract-template.md).- Required sections: Original Request, Purpose, Success Criteria (checkboxes; mark required vs optional), Constraints, Evidence Required, Iteration Log, Goal Satisfaction matrix.
- Before work: read the active goal and map planned work to saved success criteria — execution serves the saved criteria, never chat memory alone.
- After execution/verification: append an Iteration Log entry — result, evidence references (
file:line, command output, report path), remaining gaps.- Review gate: emit a Goal Satisfaction matrix —
| Success Criterion | Evidence | Status |with PASS/FAIL/BLOCKED. Overall PASS requires every required criterion PASS.- Loop rule (retry): required criterion FAIL → validate the gap is real → fix → re-review only the affected criteria. Stop cleanly when all required criteria PASS.
- Escalation rule (stop): two consecutive iterations with no criterion progressing, or a blocker needing user input → mark the criterion BLOCKED with a user-facing reason and escalate. NEVER loop indefinitely.
- Skip rule: tiny conversational tasks may skip the goal file ONLY with a recorded one-line reason. User-accepted gate skips are recorded in the goal file with reason and scope.
- Security: NEVER store secrets, tokens, credentials, or private customer data in goal files — store evidence references and redact sensitive values.
Blocked until: active goal resolved (or skip reason recorded) · saved success criteria read before edits · iteration evidence appended after execution · Goal Satisfaction matrix emitted before any PASS verdict.
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.
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.
Trade-Off Interrogation Gate — ALWAYS ask these THREE questions before ANY verdict, score, finding, or recommendation — about the thing under review AND about every recommendation YOU make. — why: naming a benefit without its price is an endorsement, not a review; the costliest trade-offs are the ones nobody wrote down.
- Is there any trade-off? Name what it SACRIFICES. "None" / "pure win" is an unfinished analysis, NOT an answer — to claim none, state which dimensions you checked and why each is unaffected: future change cost · complexity · performance/latency · memory/cost · coupling · reversibility · migration burden · operational load · blast radius · security posture · testability · team skill/ramp · delivery time · UX.
- Is it worth it? Weigh gain against sacrifice EXPLICITLY — what is gained (with a metric) · what it costs · WHO pays · WHEN it comes due — then emit WORTH IT / NOT WORTH IT / UNCLEAR. "Better" with no metric and no cost FAILS this question. NOT WORTH IT → withdraw or replace the recommendation, never keep it as-is.
- Is the trade-off material enough to CONFIRM WITH THE USER? A material trade-off is the user's call, never yours. MATERIAL when ANY holds: irreversible / one-way door (data migration, public contract, storage format, vendor lock-in) · cost shifted onto someone else (another team, ops/on-call, future maintainer, end user) · one quality attribute traded for another (correctness↔speed, security↔convenience, latency↔cost, simplicity↔flexibility) · a boundary crossed (client↔server tier, service contract, event contract, shared library) · a high-consequence path (auth, money, data integrity, breaking change, High/Medium residual risk) · the worth-it verdict is UNCLEAR.
MATERIAL → STOP and confirm via
AskUserQuestionBEFORE the verdict stands — state the trade-off, both options, what each sacrifices, and your recommendation. NOT material → record it inline with a one-line justification and proceed.Non-asking execution contexts — ESCALATE BY HANDOFF, never by silence.
AskUserQuestionreaches only the main interactive agent: a sub-agent cannot ask the user, and a terminal/verdict-only mode asks nothing by design. When you are running in such a context, the obligation is redirected, never waived — do ALL of: (a) complete questions 1 and 2 normally; (b) decide materiality and record it in the Trade-Off Assessment row withconfirmed? = NO — cannot ask from this context; (c) name the unconfirmed MATERIAL trade-off explicitly in your returned summary/verdict so the CALLER (or parent orchestrator) escalates it viaAskUserQuestionon your behalf — a material trade-off mentioned only inside a report file on disk is NOT a handoff; (d) do not emit an unqualified PASS — mark the verdict as carrying an unconfirmed material trade-off, so the caller's gate stays closed until the user answers. The caller inherits the escalation duty the moment it reads your return.This carve-out is about reachability, not convenience: it applies ONLY where the tool genuinely cannot reach the user (spawned sub-agent, terminal validate/verdict-only mode, non-interactive/headless run). It is NEVER a licence to skip the question, to self-approve a one-way door, or to downgrade materiality because asking is inconvenient — if you CAN ask, you MUST ask.
Emit a Trade-Off Assessment row per reviewed decision and per recommendation:
| decision | sacrifices | gain (metric) | who pays, when | WORTH IT/NOT/UNCLEAR | material? | confirmed? |.BLOCKED until: trade-off named (or dimensions-checked justification given) · worth-it verdict emitted · materiality decided · every MATERIAL trade-off either confirmed with the user OR — in a non-asking context — handed off in the returned verdict for the caller to confirm. A MATERIAL trade-off that is neither confirmed nor handed off can NEVER be PASS, and NEVER gets buried as a Low-severity note.
NEVER answer "no trade-off" without checking · decide a material trade-off silently on the user's behalf · let convergence/delivery pressure authorize walking through a one-way door · bundle several material trade-offs into one vague "proceed?".
Severity Rubric — Classify every finding by consequence, not by effort, reviewer preference, or how annoying the fix is. One scale applies to every review, skill, agent, workflow, and host so a tier has the same meaning everywhere. Choose the highest credible consequence supported by evidence; do not lower a tier to make a round pass.
Finding vs observation (required): An observation becomes a finding only when it names the affected user/system/data/contract, the shipped consequence, the evidence location, and the normalized tier.
INFO, advice, preference, duplicate wording, or an unsubstantiated concern is not a finding and must not reopen a loop. If the concern might affect a required behavior or gate but evidence is incomplete, emitNOT VERIFIABLEwith the missing evidence and keep it unresolved; never silently convert uncertainty into LOW.
Severity Action Definition and examples CRITICAL Block immediately; escalate Immediate material risk if shipped: authentication/authorization or safety bypass; secrets/PII exposure; irreversible destructive action; data loss/corruption; or a silent failure on a critical path. A failed binary gate that makes the result untrustworthy is represented as a separate synthetic blocker by the executable policy (not as an ordinary severity judgment). HIGH Must fix before PASS/merge Material correctness or contract risk: wrong behavior on a supported path; violated business/data invariant; meaningful privacy or authority gap; breaking API/schema/compatibility change; likely harm to users/downstream systems; or a missing proof for a behavior-changing fix. MEDIUM Must clear the current round; escalate if the fix needs an owner decision Bounded but consequential risk: an edge case, resilience/observability/testability/maintainability gap, credible future defect, or local architectural drift whose impact is real but not immediate material loss. An explicit follow-up records the escalation/residual risk; it does not make an open MEDIUM a clean pass. LOW Record and defer; never open another fix/re-review round from round 2 onward Non-blocking polish with no credible present correctness, security, privacy, authority, availability, or data-integrity impact: wording/formatting, minor documentation or convention drift, optional defensive cleanup, or a cosmetic/refinement suggestion. Consequence decision tree (apply in order): (1) Is a binary gate failed? Keep it as a separate hard blocker (the executable helper represents it as synthetic CRITICAL); do not use the ordinary severity label to hide what failed. Otherwise, would shipping permit immediate material security/safety/authority harm, irreversible destruction, data loss/corruption, or a critical-path silent failure? → CRITICAL. (2) Otherwise, does a supported path, invariant, public contract, privacy/authority boundary, compatibility promise, or behavior-changing proof fail with material user/downstream impact? → HIGH. (3) Otherwise, is there a bounded but consequential edge, resilience, observability, testability, maintainability, or architectural gap with a credible impact? → MEDIUM. (4) Otherwise, is the evidence sufficient to show only non-blocking polish with no credible present material impact? → LOW. (5) If the evidence needed to choose between steps 1–4 is missing, → NOT VERIFIABLE, not LOW. When multiple tiers fit, select the highest credible consequence; effort, implementation cost, reviewer discomfort, frequency alone, and proximity to the round cap never decide the tier.
Boundary examples (normalize before applying the round predicate): an auth bypass, exposed secret/PII, destructive command without an authority gate, or failed required test/generation/parity gate is CRITICAL; a wrong supported response, broken invariant/API/schema, meaningful privacy/authority defect, or unproven behavior-changing fix is HIGH; a bounded retry/timeout/alert/testability gap or credible maintainability drift is MEDIUM; a typo, formatting inconsistency, optional cleanup, or cosmetic suggestion proven not to affect present behavior is LOW. A missing fact about any of those boundaries is NOT VERIFIABLE until evidence or an explicitly documented residual-risk decision exists.
Classification procedure (required for every finding): (1) state the affected user, system, data, contract, or gate; (2) assess consequence if the issue ships; (3) assess exposure/likelihood and reversibility/detectability; (4) select the highest tier justified by those facts; (5) cite
file:lineor equivalent evidence and a confidence percentage. Effort, implementation cost, reviewer discomfort, and proximity to the round cap are never severity inputs.NOT VERIFIABLEis a pending evidence state, not one of the four tiers and never a LOW escape hatch: if the unresolved claim could affect required behavior, security, privacy, authority, availability, data integrity, or a binary gate, it remains an open evidence blocker until resolved or explicitly owner-accepted with documented residual risk. Classify an item LOW only when evidence supports the absence of credible present material impact.Hard-gate rule: Binary gates (tests, required artifacts, security must-fix checks, generated parity, policy compliance) are not ordinary severity-rated findings. The executable helper records a failed gate as a synthetic CRITICAL blocker solely so one predicate can carry it; the report must still name the gate and failure evidence. A failed gate blocks at every round, including when all ordinary findings are LOW; never disguise a failed gate as LOW.
Score-based skills map their numeric scale onto these tiers — do not invent a parallel vocabulary:
- 0-2 criterion scoring (e.g. production-readiness-review):
0= CRITICAL/HIGH (criterion unmet, blocks readiness),1= MEDIUM (partial, consequential gap),2= pass (no finding). If the criterion is only polish, use LOW rather than forcing a0.- Two-axis scoring (e.g. performance-review, impact × likelihood): high impact + high exposure → CRITICAL/HIGH; material impact with bounded exposure → HIGH/MEDIUM; low impact and low exposure → LOW. Record the axes and why the selected tier is the highest credible consequence.
- Scorecards /
/20grades (e.g. architecture-scalability-review): the aggregate score and verdict band are separate from finding severity. A sub-80 area is evidence to investigate, not an automatic CRITICAL/HIGH/MEDIUM/LOW label; classify each underlying gap by the consequence decision tree and keep advisory score deductions separate from blocking findings.Domain-vocabulary normalization (mandatory): Specialized skills may keep a local reporting vocabulary, but it MUST feed this same four-tier round predicate — never a second severity system:
BLOCKED,HARD FAIL, orFAILis a blocking local verdict, not an automatic CRITICAL label. Classify the underlying consequence as CRITICAL when it is an immediate material risk or failed binary gate; otherwise classify it as HIGH or MEDIUM with evidence, while preserving the local block until the owning gate is satisfied.WARNis not permission to ignore a finding. Map it to MEDIUM when the gap is consequential, to LOW only when evidence supports no credible present material impact, or upward to HIGH/CRITICAL when the consequence warrants it.PASS/compliant is not a finding.- UI
P0/P1/P2/P3/P4map to CRITICAL/HIGH/MEDIUM/LOW/LOW respectively as a starting point; override upward only when the evidence shows a higher shipped consequence. A P0/P1 accessibility or task-completion floor remains a blocking gate even when a local UI report calls it a priority rather than a severity.- Numeric SRE/readiness or impact/likelihood scores are evidence inputs, not replacement tiers. Emit the score, the consequence, and the normalized CRITICAL/HIGH/MEDIUM/LOW tier together.
INFO/advisory observations are not findings unless the evidence shows a material consequence.A finding's tier drives the gate: CRITICAL/HIGH/MEDIUM remain actionable and blocking under the round policy; LOW may be tracked as a follow-up and, from round 2, does not by itself justify another fix/re-review. An owner decision may explain or schedule an open MEDIUM but does not turn it in
…(truncated)