[Documentation] Use when scanning the codebase to (re)generate ONE project-reference doc. Parameterized by `--target=<key>`: project-structure | backend-patterns | frontend-patterns | scss-styling | design-system | code-review-rules | domain-entities | feature-spec | docs-index | e2e-tests | integration-tests | seed-test-data | ui-system. Shared 4-phase scan engine; per-target detail lives in references/targets.md. `ui-system` is an orchestrator meta-target (`kind: orchestrator`) that runs the 3 UI child scans in parallel instead of the 4-phase engine.
[Documentation] Use when scanning the codebase to (re)generate ONE project-reference doc. Parameterized by `--target=<key>`: project-structure | backend-patterns | frontend-patterns | scss-styling | design-system | code-review-rules | domain-entities | feature-spec | docs-index | e2e-tests | integration-tests | seed-test-data | ui-system. Shared 4-phase scan engine; per-target detail lives in references/targets.md. `ui-system` is an orchestrator meta-target (`kind: orchestrator`) that runs the 3 UI child scans in parallel instead of the 4-phase engine.
Codex compatibility note:
Invoke repository skills with $skill-name in Codex; this mirrored copy rewrites legacy Claude /skill-name references.
Task tracker mandate: BEFORE executing any workflow or skill step, create/update task tracking for all steps and keep it synchronized as progress changes.
User-question prompts mean to ask the user directly in Codex.
Ignore Claude-specific mode-switch instructions when they appear.
Strict execution contract: when a user explicitly invokes a skill, execute that skill protocol as written.
Subagent authorization: when a skill is user-invoked or AI-detected and its protocol requires subagents, that skill activation authorizes use of the required spawn_agent subagent(s) for that task.
Do not skip, reorder, or merge protocol steps unless the user explicitly approves the deviation first.
For workflow skills, execute each listed child-skill step explicitly and report step-by-step evidence.
If a required step/tool cannot run in this environment, stop and ask the user before adapting.
Codex Project-Reference Loading (No Hooks)
Codex uses static project-reference loading instead of runtime-injected project docs.
When coding, planning, debugging, testing, or reviewing, open project docs explicitly using this routing.
Always read:
docs/project-config.json (project-specific paths, commands, modules, and workflow/test settings)
docs/project-reference/docs-index-reference.md (routes to the full docs/project-reference/* catalog)
docs/project-reference/lessons.md (always-on guardrails and anti-patterns)
Missing/stale context route: If docs/project-config.json, the docs index, lessons.md, CLAUDE.md, AGENTS.md, or any task-required reference doc is missing or stale, auto-run $project-init or the narrow setup route ($project-config, $docs-init, $scan-all, $scan --target=<key>, $claude-md-init) before ordinary project-specific work. If Codex mirrors or AGENTS.md are missing/stale, ask the user to run $sync-codex; do not auto-run it.
Spec authoring, docs/specs/ pathing, or TC format: feature-spec-reference.md, spec-system-reference.md, spec-principles.md
Behavior/public-contract changes or spec-test-code sync: workflow-spec-test-code-cycle-reference.md plus the spec docs above
Derived spec indexes/ERDs/reimplementation guides: spec-system-reference.md and source Feature Specs under docs/specs/
Integration test implementation/review: integration-test-reference.md
E2E test implementation/review: e2e-test-reference.md
Code review/audit work: code-review-rules.md plus domain docs above based on changed files
Do not read all docs blindly. Start from docs-index-reference.md, then open only relevant files for the task.
Quick Summary
Goal: Scan the codebase for ONE target reference doc and surgically (re)populate it with actual patterns — every example from real project files with file:line. The 4-phase engine below is shared; the per-target data (which doc, how many sub-agents, what to detect, what sections to write, what NOT to do) comes from the target's entry in references/targets.md.
Workflow:
Resolve target — Read --target=<key>; load its entry from references/targets.md
Assess — Read target doc, detect init vs sync (vs force) mode, run the target's Phase-0 detection table(s)
Scan — Launch the target's sub-agents in parallel; discover patterns with file:line evidence
Report — Write structured findings to report file (incremental, not batched)
Generate — Surgical update of the reference doc from report (apply target's Target Sections + Content Rules)
Verify — Multi-round fresh-eyes review validates examples and coverage; then prompt-enhance the doc
Key Rules:
MUST ATTENTION resolve --target FIRST and load its manifest entry — every target-specific behavior (doc path, sub-agent count/roles, Phase-0 tables, Target Sections, Content Rules, special gates, anti-rationalization rows) comes from that entry, NOT from memory
MUST ATTENTION detect framework/type FIRST (per the target's Phase-0 table) — scan strategy derives from detection, never hardcoded
MUST ATTENTION every code example from actual project files with file:line — NEVER fabricate
MUST ATTENTION run graph command on key files before concluding — grep finds text, graph finds structure
Surgical update only — NEVER rewrite entire doc, NEVER remove a section without evidence it's obsolete
Some targets OVERRIDE shared output rules or add a branch (e.g. feature-spec intentionally includes directory trees; design-system has an init-mode Authoring branch with a sentinel-removal step). Always honor the target entry's "Content Rules / exceptions" and "Special slivers".
Scan (parameterized reference-doc scanner)
Phase 0.0: Resolve Target (BLOCKING — do this before anything else)
Parse --target=<key> from the invocation (e.g. $scan --target=backend-patterns). Accept the key with or without the --target= prefix.
If no target is supplied or the key is unknown → STOP and list the valid keys (see frontmatter / references/targets.md), ask the user which target to scan.
Read the target's entry in references/targets.md. That entry is the single source of truth for this run and supplies:
doc — the reference doc path this scan writes
description — the doc's purpose blurb
sub-agents — exact count + role of each parallel sub-agent
Phase 0 detection — the classification table(s) and BLOCKING gates for this target
Anti-Rationalization rows — target-specific evasions to refuse
prompt-enhance — the final $prompt-enhance <doc> step
Orchestrator branch (BLOCKING check): if the loaded entry is marked kind: orchestrator (e.g. ui-system), it does NOT run the 4-phase doc engine. SKIP Phases 0–4 below and instead follow the entry's Orchestration Procedure (pre-flight gate → launch the child --target= scans in parallel → verify each child doc has real content → summarize). Standard (single-doc scanner) targets ignore this step and continue with the shared engine below.
Everything below is the SHARED engine (standard single-doc scanner targets). Wherever it says "the target entry," read the loaded manifest entry — do not assume values from another target. Orchestrator-kind targets do not use this engine — they run their entry's Orchestration Procedure instead.
Phase 0: Classify & Assess
Before any other step, run in parallel:
Read the target's doc.
Detect mode: Init (placeholder — headings only / sentinel present) or Sync (populated). Some targets add a Force mode (user says "rebuild"/"reset" → treat as Init even if the doc exists) — honor it if the target entry defines it.
In Sync mode: list already-documented sections → skip re-scanning those unless staleness suspected.
Run the target entry's Phase 0 detection table(s) — detect framework / system type / architecture exactly as that table specifies. This is BLOCKING: grep terms and sub-agent scope derive from detection.
Load relevant paths from docs/project-config.json (e.g. contextGroups/modules/designSystem/e2eTesting/integrationTestVerify) if the target entry references them.
Run a graph command on the primary entry point: python .claude/scripts/code_graph trace <entry-file> --direction both --json (when .code-graph/graph.db exists).
Evidence gate: Confidence <60% on the target's primary detection axis → report uncertainty, DO NOT proceed with detection-specific scanning (or fall back exactly as the target entry's evidence-gate instruction specifies, e.g. "proceed with Agent 1 only").
Phase 1: Plan Scan Strategy
From the detected framework/type, derive the concrete patterns to search (naming conventions, base classes, config locations). NEVER assume these — derive from actual file evidence.
Create task tracking entries for each sub-agent listed in the target entry and for each phase before proceeding.
Phase 2: Execute Scan (Parallel Sub-Agents)
Launch the N general-purpose sub-agents defined in the target entry (count + roles vary per target — e.g. backend-patterns/domain-entities use 4, project-structure/frontend-patterns/design-system/code-review-rules/e2e-tests use 3, scss-styling/feature-spec/integration-tests use 2, docs-index uses a single main-agent scan + a fresh-eyes verifier). Give each sub-agent its Think scope + scan-target bullets verbatim from the entry. Each sub-agent MUST:
Write findings incrementally after each file/section — NEVER batch at end
Cite file:line for every pattern example
Confidence: >80% document as pattern; 60-80% document as "observed (unverified)"; <60% omit
All findings → plans/reports/scan-{target}-{YYMMDD}-{HHMM}-report.md.
Honor any conditional / ordered sub-agents from the entry (e.g. an Anti-Pattern agent that runs AFTER the discovery agents; a Cross-Service agent that runs ONLY for microservices; a BDD agent that runs ONLY if a BDD framework is detected). Honor any CRITICAL security flag the entry defines (e.g. hardcoded credentials).
Phase 3: Analyze & Generate
Read the full report. Apply the fresh-eyes protocol:
Round 1 (main agent): Build section drafts from report findings, using the target entry's Target Sections + Content Rules / exceptions.
Round 2 (fresh sub-agent, zero memory of Round 1): Sub-agent re-reads report + draft doc independently and checks (apply the target entry's Round-2 verification specifics):
Does every code example match an actual existing file (Glob verify)?
Do class/token/variable names in examples match actual declarations (Grep verify)?
Are required sections (Anti-Patterns / Coverage Report / Gap Analysis / M1-M2 Compliance / etc. as the target mandates) populated?
Coverage gaps: which Target Sections have no examples?
Round 3 only if Round 2 finds issues. Max 3 rounds → escalate to user if unresolved. (Clean Round 1 ends the scan; fresh-eyes is mandatory only after issues are found and fixed.)
Authoring branch (init mode): if the target entry defines one (e.g. design-system authors the canonical doc + token .scss), follow it exactly — including any sentinel removal (e.g. "First: REMOVE PLACEHOLDER_MARKER_SCSS") and regen-marker prepend.
Phase 4: Write & Verify
Write the updated doc with <!-- Last scanned: YYYY-MM-DD --> at top.
Surgical update only — preserve sections with no staleness, update only diverged sections; preserve manual annotations.
Verify (Glob check): ALL code example file paths exist — not just a sample of 5.
Verify (Grep check): class/token/variable names in examples match actual declarations.
Verify any target-mandated section is real, not hypothetical (Anti-Patterns / Coverage gaps / M1-M2 leaks / ports-from-config / etc.).
Run a graph command on 2-3 key files to validate call-chain accuracy.
Output-rule overrides: apply the target entry's "Content Rules / exceptions" — e.g. feature-spec intentionally INCLUDES a directory tree (overriding the shared no-trees rule); docs-index intentionally OUTPUTS glob-verified counts (its counts are the deliverable); e2e-tests/integration-tests forbid hardcoded counts and use grep-expression statistics.
Final Step: Enhance Scanned Doc (MANDATORY)
MUST ATTENTION after the doc is written and verified, create a REQUIRED final todo task and run $prompt-enhance <the target entry's doc> — why: this reference doc is injected into AI context; attention-anchoring (top/bottom Goal, inline READ summaries, token density) directly raises downstream AI output quality. A scan is NOT complete until its doc is prompt-enhanced.
task tracking (required, last task):Run $prompt-enhance <target doc> on the scanned doc
[IMPORTANT] Use task tracking to break ALL work into small tasks BEFORE starting — including tasks per file read. Prevents context loss from long files. Simple tasks: ask user whether to skip.
Prerequisites:MUST ATTENTION READ before executing:
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.
Scan & Update Reference Doc — Surgical updates only, never full rewrite.
Read existing doc first — understand current structure and manual annotations
Detect mode: Placeholder (only headings, no content) → Init mode. Has content → Sync mode.
Scan codebase for current state (grep/glob for patterns, counts, file paths)
Diff findings vs doc content — identify stale sections only
Update ONLY sections where code diverged from doc. Preserve manual annotations.
Update metadata (date, counts, version) in frontmatter or header
NEVER rewrite entire doc. NEVER remove sections without evidence they're obsolete.
Output Quality — Token efficiency without sacrificing quality.
No inventories/counts — AI can grep | wc -l. Counts go stale instantly
No directory trees — AI can glob/ls. Use 1-line path conventions
No TOCs — AI reads linearly. TOC wastes tokens
No examples that repeat what rules say — one example only if non-obvious
Lead with answer, not reasoning. Skip filler words and preamble
Sacrifice grammar for concision in reports
Unresolved questions at end, if any
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. Before changing a constant, limit, flag, wording, or pattern, read nearby context and history.
Surface ambiguity before acting — don't pick silently. Multiple valid interpretations require an explicit question or stated assumption with risk.
Keep shared guidance role-relevant. Universal guidance must help every receiving skill or agent; code-specific obligations belong only in code-specific protocols.
IMPORTANT MUST ATTENTION read existing doc first, scan codebase, diff, surgical update only. Never rewrite entire doc.
IMPORTANT MUST ATTENTION output quality: no counts/trees/TOCs, 1 example per pattern, lead with answer. (Per-target exceptions in the manifest entry override this — e.g. feature-spec trees, docs-index counts.)
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.
Closing Reminders
IMPORTANT MUST ATTENTION resolve --target and load its manifest entry FIRST — never scan from memory of "what a backend/frontend/design scan does"
IMPORTANT MUST ATTENTION — Protocols in force (concise digest of the SYNC/shared blocks this skill carries):
Critical Thinking: traced file:line proof per claim; confidence >80% to act.
Scan & Update Doc: read existing doc, diff, surgical update only — never full rewrite.
Output Quality: no counts/trees/TOCs; 1 example per pattern; lead with answer.
AI Mistake Prevention: verify generated content against evidence, trace downstream references, verify all affected outputs, re-read after context loss, surface ambiguity.
IMPORTANT MUST ATTENTION Final Step: run $prompt-enhance <target doc> as the REQUIRED last todo task — never end the scan without enhancing the doc it just wrote
IMPORTANT MUST ATTENTION break work into small task tracking tasks BEFORE starting — one task per sub-agent, one per phase
IMPORTANT MUST ATTENTION detect framework/type FIRST in Phase 0 — all grep terms derive from detection, never hardcoded
IMPORTANT MUST ATTENTION cite file:line for every pattern (confidence >80% to document; <60% omit)
IMPORTANT MUST ATTENTION run graph command on key files — grep finds text, graph finds structure (callers, event chains, blast radius)
IMPORTANT MUST ATTENTION sub-agents write findings incrementally after each file — NEVER batch at end (context loss)
IMPORTANT MUST ATTENTION read existing doc FIRST, diff findings, surgical update only — NEVER rewrite entire doc
IMPORTANT MUST ATTENTION multi-round fresh-eyes review — main agent rationalizes its own mistakes; Round 2 sub-agent catches what main agent dismissed
IMPORTANT MUST ATTENTION honor the target entry's Content-Rule exceptions, Special slivers, and Anti-Rationalization rows — they encode why this target differs from the others
Anti-Rationalization (shared — the target entry adds its own rows):
Evasion
Rebuttal
"I know what a <target> scan does, skip the manifest entry"
The entry holds the BLOCKING gates, sub-agent count, and exceptions — scanning from memory drops them
[WORKFLOW-EXECUTION-PROTOCOL] [BLOCKING] Workflow Execution Protocol — MANDATORY IMPORTANT MUST CRITICAL. Do not skip for any reason.
Generic portability boundary: Reusable skills and protocol text stay project-neutral; project-specific conventions are discovered from docs/project-config.json and docs/project-reference/. Apply shared AI-SDD from shared/sdd-artifact-contract.md. Read docs/project-config.json and docs/project-reference/docs-index-reference.md, then open the project reference docs named there. For spec, test-case, behavior-change, public-contract, or docs/specs/ work, route through the local spec docs named by the docs index: feature-spec-reference.md, spec-system-reference.md, spec-principles.md, and workflow-spec-test-code-cycle-reference.md when specs/tests/code must stay synchronized. If either file or a required reference doc is missing or stale, auto-run $project-init (or the narrow lower-level route such as $project-config, $docs-init, $scan-all, or $scan --target=<key>) before ordinary project-specific work. Any supported AI tool may execute when this shared context and local docs are available.
DETECT: If the prompt starts with an explicit slash skill/workflow command, execute it directly. Otherwise match the prompt against the workflow catalog and skill list.
ANALYZE: Choose the best option: execute directly, invoke a skill, activate a standard workflow, or compose a custom step combination.
AUTO-SELECT: Pick the best option yourself. Do not ask the user to choose between direct execution, skill, standard workflow, or custom workflow.
ACTIVATE: For a selected workflow, call $start-workflow <workflowId>; for a selected skill, invoke that skill; for a custom workflow, sequence custom steps directly; for direct execution, proceed with the task.
CREATE TASKS: task tracking for ALL workflow/skill/custom steps before execution when the selected path has multiple steps.
EXECUTE: Advance per the Workflow Step Advancement & Parallel Phases rule in your context instructions — model-driven; a sub-agent completion advances a step identically to an inline call; a parallel-phase group is an all-return barrier (advance only after ALL members return, never serialize it)
Trace every requirement or invariant through decision, task, TC/test, source evidence, and docs/spec update.
Treat code-to-spec extraction as reference-only until accepted by the canonical spec owner.
Any supported AI tool may plan, implement, review, or verify with synced context; using multiple tools is optional.
Update .claude source first, then sync generated mirrors; do not manually edit .agents, .codex, or AGENTS.md. — why: mirrors are generated artifacts; hand-edits are overwritten on the next sync
If docs/project-config.json, root instruction files, or a required project-reference doc is missing or stale, auto-run $project-init or the narrow lower-level route before ordinary project-specific work.
Active reference:shared/sdd-artifact-contract.md in the active skills root.
SYNC:ai-sdd-artifact-contract:reminder
MANDATORY Apply shared/sdd-artifact-contract.md; keep reusable AI-SDD in .claude and local rules in project docs.
MANDATORY Code-to-spec extraction is reference-only until canonical acceptance; any supported AI tool may execute with synced context.
MANDATORY Update .claude source before syncing generated mirrors; do not manually edit .agents, .codex, or AGENTS.md.
MANDATORY Missing or stale project config, root instruction files, or required reference docs route project-specific work through $project-init or the narrow setup route automatically.
[TASK-PLANNING] [MANDATORY] BEFORE executing any workflow or skill step, create/update task tracking for all planned steps, then keep it synchronized as each step starts/completes.
[LESSON-LEARNED-REMINDER] [BLOCKING] Task Planning & Continuous Improvement — MANDATORY. Do not skip.
Break work into small tasks (task tracking) before starting. Add final task: "Analyze AI mistakes & lessons learned".
Extract lessons — ROOT CAUSE ONLY, not symptom fixes:
Name the FAILURE MODE (reasoning/assumption failure), not symptom — "assumed API existed without reading source" not "used wrong enum value".
Generality test: does this failure mode apply to ≥3 contexts/codebases? If not, abstract one level up.
Write as a universal rule — strip project-specific names/paths/classes. Useful on any codebase.
Consolidate: multiple mistakes sharing one failure mode → ONE lesson.
Recurrence gate: "Would this recur in future session WITHOUT this reminder?" — No → skip $learn.
BOTH gates pass → ask user to run $learn.
[CRITICAL-THINKING-MINDSET] Apply critical thinking, sequential thinking. Every claim needs traced proof, confidence >80% to act.
Anti-hallucination principle: 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 Attention principle (Primacy-Recency): Put the 3 most critical rules at both top and bottom of long prompts/protocols so instruction adherence survives long context windows.
Goal-driven execution: Define success criteria first, loop until verified, and stop only when observable checks pass.
Tests verify intent: Tests must protect business rules/invariants and fail when the protected intent breaks, not only mirror current behavior.
Common AI Mistake Prevention (System Lessons)
Re-read files after context compaction. Edit requires prior Read in same context; compaction wipes read state. Re-read before editing.
Grep for old terms after bulk replacements. AI over-trusts find/replace completeness. Grep full repo after bulk edits for missed refs in docs/configs/catalogs.
Check downstream references before deleting. Deletions cascade doc/code staleness. Map referencing files before removal.
After memory loss, check existing state before creating new. Compaction wipes prior-work memory. Query current state to resume — never blindly duplicate.
Verify AI-generated content against actual code. AI hallucinates APIs, class names, method signatures. Grep to confirm existence before documenting/referencing.
Trace full dependency chain after edits. Changing a definition misses downstream consumers. Trace the full chain.
When renaming, grep ALL consumer file types. Some file types silently ignore missing refs (no compile error). Search code, templates, configs, generated files.
Trace ALL code paths when verifying correctness. Code existing ≠ code executing. Trace early exits, error branches, conditional skips — not just happy path.
Update docs that embed canonical data when source changes. Docs inlining derived data (workflows, schemas, configs) go stale silently. Update all embedding docs alongside source.
Verify sub-agent results after context recovery. Background agents may finish while parent compacted — grep-verify output, don't trust assumed completion.
Cross-check full target list against sub-agent assignments. Parallel sub-agents by category miss boundary items. Reconcile union of assignments against target list before proceeding.
Sub-agents inherit knowledge only from their agent .md definition — use custom agent types, not built-in Explore. Tool adoption = permission + knowledge + enforcement (numbered workflow step).
Persist sub-agent findings incrementally, not as a final batch. Long sub-agents hit cutoffs before final write — findings lost. Instruct append-per-section to report file.
When debugging, ask "whose responsibility?" before fixing. Trace caller (wrong data) vs callee (wrong handling). Fix at responsible layer — never patch symptom site.
Grep ALL removed names after extraction/refactoring. Primary file "done" ≠ secondary files clean. Grep entire scope for every removed symbol before declaring complete.
Assume existing values are intentional — ask WHY before changing. Pattern-matching as "wrong" skips context. Before changing any constant/limit/flag: read comments, git blame, surrounding code.
Verify ALL affected outputs, not just the first. One build green ≠ all green. Multi-stack changes (backend/frontend/tests/docs) require verifying EVERY output.
Evaluate fit before copying a nearby pattern. Closest example ≠ matching preconditions — verify the new context shares the same constraints, base classes, scope, lifetime.
Holistic-first debugging — resist nearest-attention trap. Don't dive into first plausible cause. List EVERY precondition (config, env vars, paths, DB, endpoints, creds, versions, DI, data). Verify each against evidence (grep/query — not reasoning). Ask "what would falsify this?" — if nothing, it's not a hypothesis. Most expensive failure: going deeper in "obvious" layer while bug sits in layer never questioned.
Surgical changes — apply the diff test (context-aware). Two modes: (1) Bug fix → every line traces to the bug; no restyling; orphan cleanup only for imports YOUR changes made unused. (2) Review/enhancement → implement improvements AND announce as "Enhancement beyond main request: [what]". Never silently scope-creep. Diff test: "Would this line exist if I wasn't asked to do X?" — if no, delete or announce.
Surface ambiguity before coding — don't pick silently. Multiple valid interpretations → present each with effort: "[Request] could mean (1) [N h], (2) [N h]. Which matters?" List scope/format/volume/constraints assumptions first. If simpler path exists, say so. Never silently pick.
[MANDATORY FIRST ACTION] ALWAYS activate a suitable skill or workflow BEFORE responding. Match task against workflow catalog + skill list; invoke via skill invocation or $start-workflow <workflowId>. NEVER answer or write code before checking. Skip = protocol violation.
Why-Review adversarial mindset — apply when reviewing any plan, decision, or design. Default SKEPTIC not VALIDATOR: steel-man a rejected alternative, invert each stated reason ("what does it sacrifice?"), stress-test top 2-3 assumptions, run pre-mortem ("ships, fails in 3 months — what breaks?"), surface 1-2 alternatives author missed. Section presence ≠ quality; quality = causal reasoning + concrete mitigations + evidence, not "it's better" or "monitor closely".
Front-load report-write in sub-agent prompts for large reviews. Many-file sub-agents hit budget before final write — findings lost. Design prompts so: (1) report-write is first explicit deliverable, (2) append per-file/section (not batched), (3) scope bounded so reads don't exhaust budget. Truncated mid-sentence with no report file → spawn narrower scope, don't retry same prompt.
After context compaction, re-verify all prior phase outcomes before continuing. Summaries describe intent, not environment state (git index, filesystem, processes). On resume, FIRST audit: git status, re-read modified files, verify filesystem. Every "completed" claim is an untested hypothesis until evidence confirms.
OOM/memory: check row count before row size. Triage: (1) Unbounded query — no DB filter for trigger? Push filter to DB; eliminates OOM. (2) Large rows? Projection reduces proportionally. Row reduction > projection in ROI.
Keep domain concepts out of generic/shared/infrastructure layers. Reusable layer (shared library, framework, infra module) must reference NO consumer-specific domain concept — tenant/customer/product IDs, business entities, feature rules. Leak compiles + runs → passes review silently while coupling the "reusable" layer to one consumer. Keep shared type domain-free; push domain fields/logic down into the consumer via subclass/composition. — why: a layer coupled to one consumer's domain is no longer reusable.
