all-for-claudecode 2.2.0 → 2.3.0

package/.claude-plugin/marketplace.json

@@ -6,14 +6,14 @@
   },
   "metadata": {
     "description": "Automated pipeline for Claude Code — spec → plan → implement → review → clean",
-    "version": "2.2.0"
+    "version": "2.3.0"
   },
   "plugins": [
     {
       "name": "afc",
       "source": "./",
       "description": "Automated pipeline for Claude Code. Automates the full development cycle: spec → plan → implement → review → clean.",
-      "version": "2.2.0",
+      "version": "2.3.0",
       "category": "automation",
       "tags": ["pipeline", "automation", "spec", "plan", "implement", "review", "critic-loop"]
     }

package/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "afc",
-  "version": "2.2.0",
+  "version": "2.3.0",
   "description": "Automated pipeline for Claude Code. Automates the full development cycle: spec → plan → implement → review → clean.",
   "author": { "name": "jhlee0409", "email": "relee6203@gmail.com" },
   "homepage": "https://github.com/jhlee0409/all-for-claudecode",

package/MIGRATION.md

@@ -1,5 +1,29 @@
 # Migration Guide
 
+## v2.3.0 — Memory Management + analyze/validate Split
+
+### What Changed
+
+| Item | Before (v2.2.x) | After (v2.3.0) |
+|------|-----------------|----------------|
+| `/afc:analyze` | Artifact consistency check (= validate) | General-purpose code analysis (user-invocable, sonnet, context: fork) |
+| `/afc:validate` | Did not exist separately | Artifact consistency check (model-only, haiku) |
+| Agent MEMORY.md | Unbounded size | 100-line limit with auto-pruning |
+| Memory directories | Unbounded file accumulation | Loading limits (most recent N files) + rotation in Clean phase |
+| `/afc:doctor` Category 6 | Hook Health | Memory Health (new); Hook Health → 7, Version Sync → 8 |
+| Checkpoint writes | Single location | Dual-write (project memory + auto-memory) |
+
+### Migration Steps
+
+**No action required for most users.** Changes are backwards-compatible and auto-applied.
+
+- If you previously used `/afc:analyze` for artifact consistency checking, use `/afc:validate` instead (or let the pipeline invoke it automatically).
+- Agent memory files (`.claude/agent-memory/afc-architect/MEMORY.md`, `.claude/agent-memory/afc-security/MEMORY.md`) exceeding 100 lines will be auto-pruned on next pipeline run.
+- Memory subdirectories exceeding rotation thresholds will be auto-pruned during the Clean phase.
+- Run `/afc:init` to update the `AFC:VERSION` tag in your CLAUDE.md.
+
+---
+
 ## v2.0 — Rebrand: selfish-pipeline → all-for-claudecode
 
 > all-for-claudecode v2.0 renames the package, plugin prefix, scripts, agents, and state files from `selfish` to `afc`.

package/README.md

@@ -56,6 +56,47 @@ CI failure → debug-based RCA (not blind retry).
 Critic Loops verify quality at each gate until convergence.
 ```
 
+## Walkthrough: What a Pipeline Run Looks Like
+
+Running `/afc:auto "Add password reset flow"` produces this (abbreviated):
+
+**Spec (1/5)** — Generates `spec.md` with requirements and acceptance criteria:
+```
+FR-001: POST /auth/reset sends email with token
+FR-002: GET /auth/reset/:token validates and shows form
+FR-003: Token expires after 1 hour
+Acceptance: Given expired token, When user submits, Then show error
+```
+
+**Plan (2/5)** — Creates `plan.md` with file change map and architecture decisions:
+```
+File Change Map:
+  src/routes/auth.ts        — ADD reset endpoint handlers
+  src/services/email.ts     — ADD sendResetEmail()
+  src/middleware/validate.ts — MODIFY add token validation
+  tests/auth.test.ts        — ADD reset flow tests
+```
+
+**Implement (3/5)** — Auto-decomposes into tasks, executes with CI gates:
+```
+Tasks: 4 total (2 parallel)
+  [1] Add reset endpoint     ✓
+  [2] Add email service       ✓  ← parallel with [1]
+  [3] Add token validation    ✓  ← depends on [1]
+  [4] Add tests               ✓
+CI: npm test → passed
+```
+
+**Review (4/5)** — 8-perspective review + specialist agent analysis:
+```
+Architecture (afc-architect): ✓ layer boundaries respected
+Security (afc-security): ⚠ rate-limit reset endpoint
+Performance: ✓ no N+1 queries
+→ Auto-fixed: added rate limiter middleware
+```
+
+**Clean (5/5)** — Removes pipeline artifacts, final CI check.
+
 ## Slash Commands
 
 | Command | Description |
@@ -76,9 +117,31 @@ Critic Loops verify quality at each gate until convergence.
 | `/afc:checkpoint` | Save session state |
 | `/afc:resume` | Restore session state |
 | `/afc:tasks` | Task decomposition (auto-generated by implement) |
-| `/afc:analyze` | Verify artifact consistency |
+| `/afc:ideate` | Explore and structure a product idea |
+| `/afc:launch` | Generate release artifacts (changelog, tag, publish) |
+| `/afc:validate` | Verify artifact consistency |
+| `/afc:analyze` | General-purpose code and component analysis |
 | `/afc:clarify` | Resolve spec ambiguities |
 
+### Individual Command Examples
+
+```bash
+# Write a spec for a specific feature
+/afc:spec "Add dark mode toggle"
+
+# Design a plan from an existing spec
+/afc:plan
+
+# Debug a specific error
+/afc:debug "TypeError: Cannot read property 'user' of undefined"
+
+# Run code review on current changes
+/afc:review
+
+# Explore and structure a product idea
+/afc:ideate "real-time collaboration feature"
+```
+
 ## Hook Events
 
 Every hook fires automatically — no configuration needed after install.
@@ -90,7 +153,7 @@ Every hook fires automatically — no configuration needed after install.
 | `PreToolUse` | Blocks dangerous commands (`push --force`, `reset --hard`) |
 | `PostToolUse` | Tracks file changes + auto-formats code |
 | `SubagentStart` | Injects pipeline context into subagents |
-| `Stop` | CI gate (shell) + code completeness check (agent) |
+| `Stop` | CI gate (shell) + code completeness check (shell) |
 | `SessionEnd` | Warns about unfinished pipeline |
 | `PostToolUseFailure` | Diagnostic hints for known error patterns |
 | `Notification` | Desktop alerts (macOS/Linux) |
@@ -103,7 +166,7 @@ Every hook fires automatically — no configuration needed after install.
 | `WorktreeCreate` | Sets up worktree isolation for parallel workers |
 | `WorktreeRemove` | Cleans up worktree after worker completion |
 
-Handler types: `command` (shell scripts, all events), `prompt` (LLM single-turn, TaskCompleted), `agent` (subagent with tools, Stop).
+Handler types: `command` (shell scripts, all events), `prompt` (LLM single-turn, TaskCompleted).
 
 ## Persistent Memory Agents
 
@@ -125,31 +188,21 @@ The implement phase automatically selects execution strategy:
 
 Dependencies are tracked via DAG. CI gate + Mini-Review + Auto-Checkpoint run at each phase boundary.
 
-## Project Presets
-
-| Preset | Stack |
-|---|---|
-| `template` | Generic (manual config) |
-| `nextjs-fsd` | Next.js + FSD + Zustand + React Query |
-| `react-spa` | Vite + React 18 + Zustand + Tailwind |
-| `express-api` | Express + TypeScript + Prisma + Jest |
-| `monorepo` | Turborepo + pnpm workspace |
-
 ## Configuration
 
 ```
 /afc:init
 ```
 
-Detects your tech stack and generates `.claude/afc.config.md` with CI/lint/test commands, architecture rules, framework settings, and code style conventions.
+Auto-detects your tech stack (package manager, framework, architecture, testing, linting) and generates `.claude/afc.config.md` with CI commands, architecture rules, and code style conventions. No manual preset selection needed — the init command analyzes your project structure directly.
 
 ## FAQ
 
 ### Does it work with any project?
-Yes. Run `/afc:init` to auto-detect your stack, or pick a preset.
+Yes. Run `/afc:init` to auto-detect your stack. Works with JavaScript/TypeScript, Python, Rust, Go, and any project with a CI command.
 
 ### Does it require any dependencies?
-No. Pure markdown commands + bash hook scripts.
+No. Pure markdown commands + bash hook scripts. No npm packages are imported at runtime.
 
 ### What happens if CI fails during the pipeline?
 Debug-based RCA: traces the error, forms a hypothesis, applies a targeted fix. Halts after 3 failed attempts with full diagnosis.
@@ -158,7 +211,22 @@ Debug-based RCA: traces the error, forms a hypothesis, applies a targeted fix. H
 Yes. Each phase has its own command (`/afc:spec`, `/afc:plan`, `/afc:implement`, `/afc:review`). `/afc:auto` runs them all.
 
 ### What are Critic Loops?
-Convergence-based quality checks after each phase. They evaluate output against criteria (completeness, feasibility, architecture compliance) and auto-fix issues until stable. 4 verdicts: PASS, FAIL, ESCALATE (asks user), DEFER.
+Convergence-based quality checks after each phase. They evaluate output against criteria and auto-fix issues until stable. 4 verdicts: PASS, FAIL, ESCALATE (asks user), DEFER.
+
+### How many tokens does a pipeline run use?
+Depends on project size and feature complexity. A typical `/afc:auto` run for a medium feature uses roughly the same as a detailed manual implementation session — the pipeline adds structure, not overhead.
+
+### Can I customize the pipeline behavior?
+Yes. Edit `.claude/afc.config.md` to change CI commands, architecture rules, and code style conventions. The pipeline reads this config at every phase.
+
+### Does it work with monorepos?
+Yes. Run `/afc:init` in the monorepo root. The init command detects workspace structure and configures accordingly.
+
+### Can multiple team members use it on the same repo?
+Yes. Each developer runs their own pipeline independently. The `.claude/afc.config.md` config is shared (commit it to the repo), but pipeline state is local and session-scoped.
+
+### How is this different from Cursor / Copilot / other AI tools?
+All-for-claudecode is not a code completion tool — it is a structured development pipeline. It enforces spec → plan → implement → review flow with quality gates, persistent memory agents, and CI verification at every step.
 
 ## License
 

package/agents/afc-architect.md

@@ -11,6 +11,8 @@ tools:
   - WebSearch
 model: sonnet
 memory: project
+# Note: no `isolation: worktree` — architect writes ADR files to project memory
+# which must persist in the main worktree (unlike afc-security which is read-only)
 skills:
   - docs/critic-loop-rules.md
   - docs/phase-gate-protocol.md
@@ -50,6 +52,11 @@ At the end of each analysis:
 1. Record new ADR decisions, discovered patterns, or architectural insights to MEMORY.md
 2. Keep entries concise — only stable patterns and confirmed decisions
 3. Remove outdated entries when architecture evolves
+4. **Size limit**: MEMORY.md must not exceed **100 lines**. If adding new entries would exceed the limit:
+   - Remove the oldest ADR entries (keep the most recent decisions)
+   - Merge similar architecture patterns into single entries
+   - Remove entries for deleted/refactored code that no longer exists
+   - Prioritize: active constraints > recent patterns > historical ADRs
 
 ## Memory Format
 

package/agents/afc-security.md

@@ -45,6 +45,12 @@ At the end of each scan:
 1. Record newly discovered vulnerability patterns to MEMORY.md
 2. Record confirmed false positives with reasoning
 3. Note project-specific security characteristics (e.g., input sanitization patterns, auth flows)
+4. **Size limit**: MEMORY.md must not exceed **100 lines**. If adding new entries would exceed the limit:
+   - Remove the oldest false positive entries (patterns likely already fixed)
+   - Merge similar vulnerability patterns into single entries
+   - Remove entries for files/paths that no longer exist in the codebase
+   - Prioritize: active vulnerability patterns > project security profile > historical false positives
+   - Never remove entries for Critical-severity patterns regardless of age
 
 ## Memory Format
 

package/commands/analyze.md

@@ -1,125 +1,103 @@
 ---
 name: afc:analyze
-description: "Artifact consistency validation (read-only)"
-argument-hint: "[validation scope: spec-plan, tasks-only]"
-user-invocable: false
+description: "General-purpose code and component analysis"
+argument-hint: "<analysis target or question>"
+user-invocable: true
 context: fork
 allowed-tools:
   - Read
   - Grep
   - Glob
-model: haiku
+  - WebSearch
+model: sonnet
 ---
 
-# /afc:analyze — Artifact Consistency Validation
+# /afc:analyze — Code Analysis
 
-> Validates consistency and quality across spec.md, plan.md, and tasks.md.
+> Performs general-purpose codebase exploration and analysis based on a natural-language prompt.
 > **Read-only** — does not modify any files.
 
 ## Arguments
 
-- `$ARGUMENTS` — (optional) limit validation scope (e.g., "spec-plan", "tasks-only")
+- `$ARGUMENTS` — (required) description of what to analyze (e.g., "trace the login flow", "root cause of rendering bug", "how does the hook system work")
 
 ## Config Load
 
 **Always** read `.claude/afc.config.md` first. This file contains free-form markdown sections:
-- `## Architecture` — architecture pattern, layers, import rules (primary reference for this command)
+- `## Architecture` — architecture pattern, layers, import rules (primary reference for structural analysis)
 - `## Code Style` — language, naming conventions, lint rules
 - `## Project Context` — framework, state management, testing, etc.
 
-If config file is missing: read `CLAUDE.md` for architecture info. Assume "Layered Architecture" if neither source has it.
+If config file is missing: read `CLAUDE.md` for architecture info. Proceed without config if neither exists.
 
 ## Execution Steps
 
-### 1. Load Artifacts
+### 1. Parse Analysis Intent
 
-From `.claude/afc/specs/{feature}/`:
-- **spec.md** (required)
-- **plan.md** (required)
-- **tasks.md** (if present)
-- **research.md** (if present)
+Classify `$ARGUMENTS` into one of these analysis modes:
 
-Warn about missing files but proceed with what is available.
+| Mode | Trigger Keywords | Focus |
+|------|-----------------|-------|
+| **Root Cause** | "why", "cause", "bug", "error", "broken" | Error trace → data flow → hypothesis |
+| **Structural** | "how", "architecture", "flow", "trace", "structure" | Component relationships, call graphs, data flow |
+| **Exploratory** | "what", "find", "where", "which", "list" | File/function discovery, pattern matching |
+| **Comparative** | "difference", "compare", "vs", "between" | Side-by-side analysis of implementations |
 
-### 2. Run Validation
+If the intent doesn't clearly match a mode, default to **Exploratory**.
 
-Validate across 6 categories:
+### 2. Codebase Exploration
 
-#### A. Duplication Detection (DUPLICATION)
-- Similar requirements within spec.md
-- Overlapping tasks within tasks.md
+Based on the classified mode:
 
-#### B. Ambiguity Detection (AMBIGUITY)
-- Unmeasurable adjectives ("appropriate", "fast", "good")
-- Residual TODO/TBD/FIXME markers
-- Incomplete sentences
+1. **Identify scope**: determine which files/directories are relevant to `$ARGUMENTS`
+2. **Read code**: read relevant files using Read tool (prioritize by relevance)
+3. **Trace connections**: follow imports, function calls, and data flow
+4. **Gather evidence**: collect specific code references (file:line) for findings
 
-#### C. Coverage Gaps (COVERAGE)
-- spec → plan: Are all FR-*/NFR-* reflected in the plan?
-- plan → tasks: Are all items in the plan's File Change Map present in tasks?
-- spec → tasks: Are all requirements mapped to tasks?
+Exploration should be guided by `{config.architecture}` layer structure when available.
 
-#### D. Inconsistencies (INCONSISTENCY)
-- Terminology drift (different names for the same concept)
-- Conflicting requirements
-- Mismatches between technical decisions in plan and execution in tasks
+### 3. Analysis
 
-#### E. Principles Compliance (PRINCIPLES)
-- Validate against MUST principles in .claude/afc/memory/principles.md if present
-- Potential violations of {config.architecture} rules
+Apply the appropriate analysis lens:
 
-#### F. Unidentified Risks (RISK)
-- Are there risks not identified in plan.md?
-- External dependency risks
-- Potential performance bottlenecks
-
-### 3. Severity Classification
-
-| Severity | Criteria |
-|----------|----------|
-| **CRITICAL** | Principles violation, core feature blocker, security issue |
-| **HIGH** | Duplication/conflict, untestable, coverage gap |
-| **MEDIUM** | Terminology drift, ambiguous requirements |
-| **LOW** | Style improvements, minor duplication |
+- **Root Cause**: build a causal chain from symptom → intermediate causes → root cause
+- **Structural**: map component relationships, identify coupling and cohesion patterns
+- **Exploratory**: enumerate findings with code references
+- **Comparative**: highlight similarities, differences, and tradeoffs
 
 ### 4. Output Results (console)
 
 ```markdown
-## Consistency Analysis Results: {feature name}
+## Analysis: {summary of $ARGUMENTS}
+
+### Mode: {Root Cause | Structural | Exploratory | Comparative}
 
 ### Findings
-| ID | Category | Severity | Location | Summary | Recommended Action |
-|----|----------|----------|----------|---------|-------------------|
-| A-001 | COVERAGE | HIGH | spec FR-003 | No mapping in tasks | Add task |
-| A-002 | AMBIGUITY | MEDIUM | spec NFR-001 | "quickly" is unmeasurable | Add numeric threshold |
-
-### Coverage Summary
-| Mapping | Coverage |
-|---------|----------|
-| spec → plan | {N}% |
-| plan → tasks | {N}% |
-| spec → tasks | {N}% |
-
-### Metrics
-- Total requirements: {N}
-- Total tasks: {N}
-- Issues: CRITICAL {N} / HIGH {N} / MEDIUM {N} / LOW {N}
-
-### Next Steps
-{Concrete action proposals for CRITICAL/HIGH issues}
+{Numbered findings with code references (file:line)}
+
+### Key Relationships
+{Relevant component/function relationships discovered}
+
+### Summary
+{2-3 sentence conclusion answering the original question}
+
+### Suggested Next Steps
+{1-3 actionable suggestions based on the analysis}
 ```
 
 ### 5. Final Output
 
 ```
-Analysis complete
-├─ Found: CRITICAL {N} / HIGH {N} / MEDIUM {N} / LOW {N}
-├─ Coverage: spec→plan {N}%, plan→tasks {N}%, spec→tasks {N}%
-└─ Recommended: {next action}
+Analysis complete: {short summary}
+├─ Mode: {mode}
+├─ Files explored: {N}
+├─ Findings: {N}
+└─ Suggested next steps: {N}
 ```
 
 ## Notes
 
 - **Read-only**: Do not modify any files. Report only.
-- **Avoid false positives**: Do not over-flag ambiguity. Consider context.
-- **Optional**: Not required in the pipeline. Can proceed plan → implement directly.
+- **Scope discipline**: Focus analysis on what was asked. Do not expand scope unnecessarily.
+- **Code references**: Always include `file:line` references so the user can navigate to relevant code.
+- **Not artifact validation**: For spec/plan/tasks consistency checks, use `/afc:validate` instead.

package/commands/architect.md

@@ -92,7 +92,7 @@ Structure analysis results and **print to console**:
 
 > **Always** read `${CLAUDE_PLUGIN_ROOT}/docs/critic-loop-rules.md` first and follow it.
 
-Run the critic loop until convergence. Safety cap: 7 passes.
+Run the critic loop until convergence. Safety cap: 7 passes (higher than the standard 5 because architecture analysis involves broader exploration across modules and layers).
 
 | Criterion | Validation |
 |-----------|------------|

package/commands/auto.md

@@ -170,7 +170,7 @@ Execute `/afc:spec` logic inline:
 4. `[NEEDS CLARIFICATION]` items: **research first, then auto-resolve remaining** (clarify skipped if Phase 0.5 already ran)
    - Items answerable via research → resolve with researched facts, tag `[RESEARCHED]`
    - Items requiring user judgment → auto-resolve with best-guess, tag `[AUTO-RESOLVED]`
-5. **Retrospective check**: if `.claude/afc/memory/retrospectives/` exists, load and check:
+5. **Retrospective check**: if `.claude/afc/memory/retrospectives/` exists, load the **most recent 10 files** (sorted by filename descending) and check:
    - Were there previous `[AUTO-RESOLVED]` items that turned out wrong? Flag similar patterns.
    - Were there scope-related issues in past specs? Warn about similar ambiguities.
 6. **Critic Loop until convergence** (safety cap: 5, follow Critic Loop rules):
@@ -191,9 +191,9 @@ Execute `/afc:plan` logic inline:
 1. Load spec.md
 2. If technical uncertainties exist → auto-resolve via WebSearch/code exploration → create research.md
 3. **Memory loading** (skip gracefully if directories are empty or absent):
-   - **Quality history**: if `.claude/afc/memory/quality-history/*.json` exists, load recent entries and display trend summary: "Last {N} pipelines: avg critic_fixes {X}, avg ci_failures {Y}, avg escalations {Z}". Use trends to inform plan risk assessment.
-   - **Decisions**: if `.claude/afc/memory/decisions/` exists, load ADR entries and check for conflicts with the current feature's design direction. Flag any contradictions.
-   - **Reviews**: if `.claude/afc/memory/reviews/` exists, scan for recurring finding patterns (same file/category appearing in 2+ reviews). Flag as known risk areas.
+   - **Quality history**: if `.claude/afc/memory/quality-history/*.json` exists, load the **most recent 10 files** (sorted by filename descending) and display trend summary: "Last {N} pipelines: avg critic_fixes {X}, avg ci_failures {Y}, avg escalations {Z}". Use trends to inform plan risk assessment.
+   - **Decisions**: if `.claude/afc/memory/decisions/` exists, load the **most recent 30 files** (sorted by filename descending) and check for conflicts with the current feature's design direction. Flag any contradictions.
+   - **Reviews**: if `.claude/afc/memory/reviews/` exists, load the **most recent 15 files** (sorted by filename descending) and scan for recurring finding patterns (same file/category appearing in 2+ reviews). Flag as known risk areas.
 4. Create `.claude/afc/specs/{feature}/plan.md`
    - **If setting numerical targets (line counts etc.), include structure-analysis-based estimates** (e.g., "function A ~50 lines, component B ~80 lines → total ~130 lines")
 5. **Critic Loop until convergence** (safety cap: 5, follow Critic Loop rules):
@@ -231,14 +231,18 @@ Execute `/afc:plan` logic inline:
    ```
    - If architect returns conflicts → **ESCALATE** to user with conflict details
    - If no conflicts → proceed (ADR recorded for future reference)
-8. **Session context preservation**: Save key decisions and constraints for context compaction resilience:
-   ```
-   save_session_context({
-     goal: { original_request: "$ARGUMENTS", current_objective: "Implement {feature}" },
-     decisions: [{ what: "{key design decision}", why: "{rationale}" }],
-     discoveries: [{ file: "{path}", insight: "{finding}" }]
-   })
+8. **Session context preservation**: Write key decisions to `.claude/afc/specs/{feature}/context.md` for compaction resilience:
+   ```markdown
+   # Session Context: {feature}
+   ## Goal
+   - Original request: $ARGUMENTS
+   - Current objective: Implement {feature}
+   ## Key Decisions
+   - {what}: {rationale}
+   ## Discoveries
+   - {file path}: {finding}
    ```
+   This file is read at Implement start to restore context after compaction.
 9. **Checkpoint**: phase transition already recorded by `afc-pipeline-manage.sh phase plan` at phase start
 10. Progress: `✓ 2/5 Plan complete (Critic: converged ({N} passes, {M} fixes, {E} escalations), files: {N}, ADR: {N} recorded, Implementation Context: {W} words)`
 
@@ -246,7 +250,7 @@ Execute `/afc:plan` logic inline:
 
 `"${CLAUDE_PLUGIN_ROOT}/scripts/afc-pipeline-manage.sh" phase implement`
 
-**Session context reload**: At implement start, call `load_session_context()` to restore key decisions and constraints from Plan phase (resilient to context compaction).
+**Session context reload**: At implement start, read `.claude/afc/specs/{feature}/context.md` if it exists. This restores key decisions and constraints from Plan phase (resilient to context compaction).
 
 Execute `/afc:implement` logic inline — **follow all orchestration rules defined in `commands/implement.md`** (task generation, mode selection, batch/swarm execution, failure recovery, task execution pattern). The implement command is the single source of truth for orchestration details.
 
@@ -255,7 +259,7 @@ Execute `/afc:implement` logic inline — **follow all orchestration rules defin
 #### Step 3.1: Task Generation + Validation
 
 1. Generate tasks.md from plan.md File Change Map (as defined in implement.md Step 1.3)
-2. **Retrospective check**: if `.claude/afc/memory/retrospectives/` exists, load and check:
+2. **Retrospective check**: if `.claude/afc/memory/retrospectives/` exists, load the **most recent 10 files** (sorted by filename descending) and check:
    - Were there previous parallel conflict issues ([P] file overlaps)? Flag similar file patterns.
    - Were there tasks that were over-decomposed or under-decomposed? Adjust granularity.
 3. Script validation (DAG + parallel overlap) — no critic loop, script-based only
@@ -410,20 +414,20 @@ Execute `/afc:review` logic inline — **follow all review perspectives defined
    - **G. Maintainability** — AI/human comprehension, naming clarity, self-contained files (direct review)
    - **H. Extensibility** — extension points, OCP, future modification cost (direct review)
 4. **Auto-resolved validation**: Check all `[AUTO-RESOLVED]` items from spec phase — does the implementation match the guess? Flag mismatches as Critical.
-5. **Past reviews check**: if `.claude/afc/memory/reviews/` exists, scan for recurring finding patterns across past review reports. Prioritize those areas.
-6. **Retrospective check**: if `.claude/afc/memory/retrospectives/` exists, load and check:
+5. **Past reviews check**: if `.claude/afc/memory/reviews/` exists, load the **most recent 15 files** (sorted by filename descending) and scan for recurring finding patterns across past review reports. Prioritize those areas.
+6. **Retrospective check**: if `.claude/afc/memory/retrospectives/` exists, load the **most recent 10 files** (sorted by filename descending) and check:
    - Were there recurring Critical finding categories in past reviews? Prioritize those perspectives.
    - Were there false positives that wasted effort? Reduce sensitivity for those patterns.
-6. **Critic Loop until convergence** (safety cap: 5, follow Critic Loop rules):
+7. **Critic Loop until convergence** (safety cap: 5, follow Critic Loop rules):
    - COMPLETENESS: were all changed files reviewed across all 8 perspectives (A-H)?
    - SPEC_ALIGNMENT: cross-check implementation against spec.md — (1) every SC verified with `{M}/{N}` count, (2) every acceptance scenario (GWT) has corresponding code path, (3) no spec constraint is violated
    - PRECISION: are there unnecessary changes? Are there out-of-scope modifications?
    - FAIL → auto-fix and continue. ESCALATE → pause, present options, resume after response. DEFER → record reason, mark clean.
-7. **Handling SC shortfalls**:
+8. **Handling SC shortfalls**:
    - Fixable → attempt auto-fix → re-run `{config.ci}` verification
    - Not fixable → state in final report with reason (no post-hoc rationalization; record as Plan-phase target-setting error)
-8. **Checkpoint**: phase transition already recorded by `afc-pipeline-manage.sh phase review` at phase start
-9. Progress: `✓ 4/5 Review complete (Critical:{N} Warning:{N} Info:{N}, SC shortfalls: {N})`
+9. **Checkpoint**: phase transition already recorded by `afc-pipeline-manage.sh phase review` at phase start
+10. Progress: `✓ 4/5 Review complete (Critical:{N} Warning:{N} Info:{N}, SC shortfalls: {N})`
 
 ### Phase 5: Clean (5/5)
 
@@ -448,7 +452,23 @@ Artifact cleanup and codebase hygiene check after implementation and review:
    - **If retrospective.md exists** → record as patterns missed by the Plan phase Critic Loop in `.claude/afc/memory/retrospectives/` (reuse as RISK checklist items in future runs)
    - **If review-report.md exists** → copy to `.claude/afc/memory/reviews/{feature}-{date}.md` before .claude/afc/specs/ deletion
    - **If research.md exists** and was not already persisted in Plan phase → copy to `.claude/afc/memory/research/{feature}.md`
-   - **Agent memory consolidation**: architect and security agents have already updated their persistent MEMORY.md during Review phase — no additional action needed (agent memory survives across sessions independently)
+   - **Agent memory consolidation**: architect and security agents have already updated their persistent MEMORY.md during Review phase. **Size enforcement**: check each agent's MEMORY.md line count — if either exceeds 100 lines, invoke the respective agent to self-prune:
+     ```
+     Task("Memory cleanup: afc-architect", subagent_type: "afc:afc-architect",
+       prompt: "Your MEMORY.md exceeds 100 lines. Read it, prune old/redundant entries, and rewrite to under 100 lines following your size limit rules.")
+     ```
+     (Same pattern for afc-security if needed. Skip if both are under 100 lines.)
+   - **Memory rotation**: for each memory subdirectory, check file count and prune oldest files if over threshold:
+     | Directory | Threshold | Action |
+     |-----------|-----------|--------|
+     | `quality-history/` | 30 files | Delete oldest files beyond threshold |
+     | `reviews/` | 40 files | Delete oldest files beyond threshold |
+     | `retrospectives/` | 30 files | Delete oldest files beyond threshold |
+     | `research/` | 50 files | Delete oldest files beyond threshold |
+     | `decisions/` | 60 files | Delete oldest files beyond threshold |
+     - Sort by filename ascending (oldest first), delete excess
+     - Log: `"Memory rotation: {dir} pruned {N} files"`
+     - Skip directories that do not exist or are under threshold
 5. **Quality report** (structured pipeline metrics):
    - Generate `.claude/afc/memory/quality-history/{feature}-{date}.json` with the following structure:
      ```json
@@ -478,7 +498,7 @@ Artifact cleanup and codebase hygiene check after implementation and review:
      ```
    - Create `.claude/afc/memory/quality-history/` directory if it does not exist
 6. **Checkpoint reset**:
-   - Clear `.claude/afc/memory/checkpoint.md` (pipeline complete = session goal achieved)
+   - Clear `.claude/afc/memory/checkpoint.md` **and** `~/.claude/projects/{ENCODED_PATH}/auto-memory/checkpoint.md` (pipeline complete = session goal achieved, dual-delete prevents stale checkpoint in either location; `ENCODED_PATH` = project path with `/` replaced by `-`)
 7. **Timeline finalize**:
    ```bash
    "${CLAUDE_PLUGIN_ROOT}/scripts/afc-pipeline-manage.sh" log pipeline-end "Pipeline complete: {feature}"
@@ -548,7 +568,7 @@ Pipeline aborted (Phase {N}/5)
 - **[P] parallel is mandatory**: if a [P] marker is assigned in tasks.md, it must be executed in parallel. Orchestration mode (batch vs swarm) is selected automatically based on task count. Sequential substitution is prohibited.
 - **Swarm mode is automatic**: when a phase has 6+ [P] tasks, the orchestrator pre-assigns tasks to swarm workers. Do not manually batch.
 - **Implementation Context travels with workers**: every sub-agent prompt includes the Implementation Context section from plan.md, ensuring spec intent propagates to parallel workers.
-- **Session context resilience**: key decisions are saved via `save_session_context` at Plan completion and restored at Implement start, surviving context compaction.
+- **Session context resilience**: key decisions are written to `.claude/afc/specs/{feature}/context.md` at Plan completion and read at Implement start, surviving context compaction.
 - **Specialist agents enhance review**: afc-architect and afc-security agents are invoked during Review to provide persistent-memory-aware analysis. Their findings are merged into the consolidated review. Agent memory updates happen automatically during the agent call.
 - **Debug-based RCA replaces blind retry**: CI failures trigger `/afc:debug` logic (hypothesis → targeted fix) instead of generic "retry 3 times". This produces better fixes and records patterns via retrospective.
 - **Acceptance tests close the spec-to-code gap**: When spec contains GWT scenarios and a test framework is configured, acceptance tests are auto-generated after implementation, verifying spec intent is met.

package/commands/doctor.md

@@ -84,7 +84,21 @@ Run ALL checks regardless of earlier failures. Do not short-circuit.
 | No lingering safety tags | `git tag -l 'afc/pre-*'` | No tags, or tags match active pipeline | ⚠ Warning: lingering safety tag `afc/pre-{x}` found. Fix: `git tag -d afc/pre-{x}` |
 | Checkpoint state | Read `.claude/afc/memory/checkpoint.md` if exists | No checkpoint (clean), or checkpoint is from current session | ⚠ Warning: stale checkpoint from {date}. Fix: run `/afc:resume` to continue or delete `.claude/afc/memory/checkpoint.md` |
 
-### Category 6: Hook Health
+### Category 6: Memory Health
+
+> Checks `.claude/afc/memory/` subdirectory sizes and agent memory file sizes. If memory directory does not exist, print `✓ No memory directory` and skip this category.
+
+| Check | How | Pass | Fail |
+|-------|-----|------|------|
+| quality-history count | Count files in `.claude/afc/memory/quality-history/` | ≤ 30 files | ⚠ Warning: {N} files in quality-history/ (threshold: 30). Oldest files should be pruned. Fix: run a pipeline with `/afc:auto` (Clean phase auto-prunes) or manually delete oldest files |
+| reviews count | Count files in `.claude/afc/memory/reviews/` | ≤ 40 files | ⚠ Warning: {N} files in reviews/ (threshold: 40). Fix: run a pipeline or manually delete oldest files |
+| retrospectives count | Count files in `.claude/afc/memory/retrospectives/` | ≤ 30 files | ⚠ Warning: {N} files in retrospectives/ (threshold: 30). Fix: run a pipeline or manually delete oldest files |
+| research count | Count files in `.claude/afc/memory/research/` | ≤ 50 files | ⚠ Warning: {N} files in research/ (threshold: 50). Fix: run a pipeline or manually delete oldest files |
+| decisions count | Count files in `.claude/afc/memory/decisions/` | ≤ 60 files | ⚠ Warning: {N} files in decisions/ (threshold: 60). Fix: run a pipeline or manually delete oldest files |
+| afc-architect MEMORY.md size | Count lines in `.claude/agent-memory/afc-architect/MEMORY.md` (if exists) | ≤ 100 lines | ⚠ Warning: afc-architect MEMORY.md is {N} lines (limit: 100). Fix: invoke `/afc:architect` to trigger self-pruning, or manually edit the file |
+| afc-security MEMORY.md size | Count lines in `.claude/agent-memory/afc-security/MEMORY.md` (if exists) | ≤ 100 lines | ⚠ Warning: afc-security MEMORY.md is {N} lines (limit: 100). Fix: invoke `/afc:security` to trigger self-pruning, or manually edit the file |
+
+### Category 7: Hook Health
 
 | Check | How | Pass | Fail |
 |-------|-----|------|------|
@@ -92,7 +106,7 @@ Run ALL checks regardless of earlier failures. Do not short-circuit.
 | All scripts exist | For each script referenced in hooks.json, check file exists | All scripts found | ✗ Fix: reinstall plugin |
 | Scripts executable | Check execute permission on each script in plugin's scripts/ | All have +x | Fix: `chmod +x` on the missing scripts, or reinstall plugin |
 
-### Category 7: Version Sync (development only)
+### Category 8: Version Sync (development only)
 
 > Only run if current directory is the all-for-claudecode source repo (check for `package.json` with `"name": "all-for-claudecode"`).
 

package/commands/implement.md

@@ -94,9 +94,9 @@ If tasks.md already exists (e.g., from standalone `/afc:tasks` run): use as-is,
    - Identify already-completed `[x]` tasks
 2. Load **Implementation Context** section from plan.md (used in sub-agent prompts)
 
-### 1.5. Retrospective Check
+### 1.7. Retrospective Check
 
-If `.claude/afc/memory/retrospectives/` exists, load and check:
+If `.claude/afc/memory/retrospectives/` exists, load the **most recent 10 files** (sorted by filename descending) and check:
 - Were there implementation issues in past pipelines (e.g., file conflicts, unexpected dependencies, CI failures after parallel execution)?
 - Flag similar patterns in the current task list. Warn before implementation begins.
 - Skip gracefully if directory is empty or absent.