AgentStack
SKILL verified MIT Self-run

Afc:tasks

skill-jhlee0409-all-for-claudecode-tasks · by jhlee0409

Task decomposition from plan with dependency tracking

No reviews yet
0 installs
16 views
0.0% view→install

Install

$ agentstack add skill-jhlee0409-all-for-claudecode-tasks

✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures

What it can access

  • Network access No
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets No
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

Are you the author of Afc:tasks? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

/afc:tasks — Task Decomposition

> Generates an executable task list (tasks.md) from plan.md. > Validates coverage with a convergence-based Critic Loop. > > Note: In /afc:auto, task generation runs automatically at implement start. Use this command only for standalone manual control.

Arguments

  • $ARGUMENTS — (optional) additional constraints or priority directives

Config Load

Read .claude/afc.config.md first. If missing: print ".claude/afc.config.md not found. Run /afc:init first." then abort.

Execution Steps

1. Load Context

From .claude/afc/specs/{feature}/:

  • plan.md (required) — if missing: "Run /afc:plan first."
  • spec.md (required), research.md (if present)

Extract from plan.md: phase breakdown, File Change Map, architecture decisions.

2. Decompose Tasks

Task Format
- [ ] T{NNN} {[P]} {[US*]} {description} `{file path}` {depends: [TXXX, ...]}

| Component | Required | Description | |-----------|----------|-------------| | T{NNN} | Yes | 3-digit sequential ID | | [P] | No | Mandatory parallel execution — no file overlap with other [P] tasks in same phase | | [US*] | No | User Story label from spec.md | | description | Yes | Verb-first clear description | | ` file path | Yes | Primary target file | | depends:` | No | Blocking dependency list |

Phase Structure
# Tasks: {feature name}
## Phase 1: Setup    — type definitions, configuration
## Phase 2: Core     — business logic, store, API
## Phase 3: UI       — components, interactions
## Phase 4: Integration & Polish — error handling, optimization
Decomposition Rules
  1. 1 task = 1 file (where possible)
  2. Same file → sequential; different files → [P] candidate
  3. [P] is mandatory, not optional — once marked, task MUST run in parallel
  4. Validate [P] file overlaps: run "${CLAUDE_SKILL_DIR}/../../scripts/afc-parallel-validate.sh" .claude/afc/specs/{feature}/tasks.md
  5. Validate DAG (no cycles): run "${CLAUDE_SKILL_DIR}/../../scripts/afc-dag-validate.sh" .claude/afc/specs/{feature}/tasks.md — abort if cycle detected
  6. Include a verification task per testable unit
  7. Add {config.gate} validation task at end of each Phase
  8. No over-decomposition: skip separate tasks for single-line changes

3. Retrospective Check

If .claude/afc/memory/retrospectives/ exists, load the 10 most recent files (sorted descending) and check for prior [P] conflict patterns or granularity issues.

4. Critic Loop

Read [docs/critic-loop-rules.md](../../docs/critic-loop-rules.md) first and follow all rules. Safety cap: 5 passes.

| Criterion | Validation | |-----------|------------| | COVERAGE | All files in plan.md File Change Map included? All FR-* in spec.md covered? | | DEPENDENCIES | DAG valid? [P] tasks in same phase have no file overlaps? All depends: IDs exist? |

Verdicts: PASS / FAIL (auto-fix, continue) / ESCALATE (pause for user) / DEFER (record, continue)

On convergence: ✓ Critic converged ({N} passes, {M} fixes, {E} escalations) On safety cap: ⚠ Critic safety cap ({N} passes). Review recommended.

5. Coverage Mapping

## Coverage Mapping
| Requirement | Tasks |
|-------------|-------|
| FR-001 | T003, T007 |
| NFR-001 | T012 |

Every FR-/NFR- must map to at least one task.

5.5. Auto-Checkpoint (standalone only)

When not inside /afc:auto, write .claude/afc/memory/checkpoint.md:

  • branch, last commit, feature name, phase (tasks complete), task count, next step (/afc:implement)

6. Final Output

Save to .claude/afc/specs/{feature}/tasks.md, then:

Tasks generated
├─ .claude/afc/specs/{feature}/tasks.md
├─ Tasks: {total} ({[P] count} parallelizable)
├─ Phases: {count}
├─ Coverage: FR {N}%, NFR {N}%
├─ Critic: converged ({N} passes, {M} fixes, {E} escalations)
└─ Next step: /afc:validate (optional) or /afc:implement

Notes

  • Cross-phase depends: prohibited: only reference tasks in same or previous phase
  • addBlockedBy auto-unblocking is guaranteed in Agent Teams only; in sub-agent mode the orchestrator must poll manually
  • Accurate file paths: use actual project structure — no guessing

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet — be the first.

Versions

  • v0.1.0 Imported from the upstream source.