Agent-Skills.md

Agent Skills: Iterate

Autonomous TDD implementation workflow with phase gates.

UncategorizedID: c-daly/agent-swarm/iterate

Repository

c-daly/agent-swarm
c-daly

Install this agent skill to your local

pnpm dlx add-skill https://github.com/c-daly/agent-swarm/tree/HEAD/skills/iterate

Skill Files

Browse the full folder contents for iterate.

Download Skill

Loading file tree…

skills/iterate/SKILL.md

Skill Metadata

Name
iterate
Description
Autonomous TDD implementation workflow with phase gates.

Iterate

Phases

test_writing → implement → test → review

| Phase | Purpose | Allowed | Blocked | |-------|---------|---------|---------| | test_writing | Write tests first (TDD) | Read, Glob, Grep, Edit, Write, Bash | - | | implement | Make tests pass | Read, Glob, Grep, Edit, Write, Bash | - | | test | Run tests + adversary gate | Read, Glob, Grep, Bash | Edit, Write | | review | Verify quality, commit+push | Read, Glob, Grep, Edit, Write, Bash | - |

Phase Flow

test_writing

  • Write tests for task requirements
  • Cover edge cases
  • No implementation yet
  • → implement

implement

  • Write code to make tests pass
  • Follow existing conventions
  • Minimal changes only
  • → test

test

  • Run pytest, ruff, coverage
  • Call adversary_gate tool (autonomous: analyzes, writes tests, runs them)
  • No manual editing — adversary tool writes directly
  • Adversary returns: pass/fail + confidence verdict

review

  • Check quality, conventions, correctness
  • clean → commit + push → done
  • issues → implement

Adversary Gate

Called during test phase. Autonomous tool that:

  1. Analyzes implementation for untested paths, edge cases, boundary conditions
  2. Writes adversarial tests directly (agent or heuristic, swappable)
  3. Runs them
  4. Returns verdict: pass/fail + confidence (weak/strong)

Implementer does not manage the adversary. Reacts to outcomes only.

Kick-back Table

| Result | → Phase | |--------|---------| | all tests pass + confident | review | | adversary tests fail | implement | | tests pass + weak suite | test_writing | | review clean | done (commit+push) | | review issues | implement |

CLI

python3 lib/iterate_workflow.py start "desc" [max_iter]
python3 lib/iterate_workflow.py status
python3 lib/iterate_workflow.py phase
python3 lib/iterate_workflow.py advance
python3 lib/iterate_workflow.py set-phase <phase>
python3 lib/iterate_workflow.py stop

Exit Conditions

  • review_approved: review clean, committed+pushed
  • max_iterations: hit limit (default: 5)
  • user_stopped: manual stop