Agent skill
autodev
Manage the PROGRAM.md operational contract that bounds autonomous development loops. Use when setting up or validating PROGRAM.md/AUTODEV.md, choosing between autodev and evolve, or turning a repeated postmortem -> analyze -> plan -> pre-mortem -> implement -> validate cycle into the v2 autonomous workflow.
Install this agent skill to your Project
npx add-skill https://github.com/boshu2/agentops/tree/main/skills-codex/autodev
SKILL.md
$autodev
$autodev manages the repo-local operational contract for autonomous
development. It does not replace $evolve or $rpi.
PROGRAM.mdorAUTODEV.mddefines the contract: mutable scope, immutable scope, experiment unit, validation commands, decision policy, escalation rules, and stop conditions.ao autodevcreates, inspects, and validates that contract.ao evolveruns the v2 autonomous improvement loop.ao rpiruns one research -> plan -> implement -> validate lifecycle.
Codex Lifecycle Guard
When this skill runs in Codex hookless mode (CODEX_THREAD_ID is set or
CODEX_INTERNAL_ORIGINATOR_OVERRIDE is Codex Desktop), ensure startup context
before editing or validating the contract:
ao codex ensure-start 2>/dev/null || true
Routing
Use this split when the user asks whether the old evolve flow should become a new command or skill:
| Intent | Action |
|---|---|
| define or repair the repo-local autonomous policy | use $autodev and ao autodev |
| run the autonomous improvement loop | use $evolve or ao evolve |
| run one bounded lifecycle | use $rpi or ao rpi |
PROGRAM.md takes precedence over AUTODEV.md. Treat AUTODEV.md as the
compatibility alias.
Execution Steps
Step 1: Detect the contract
if [ -f PROGRAM.md ]; then
PROGRAM_PATH=PROGRAM.md
elif [ -f AUTODEV.md ]; then
PROGRAM_PATH=AUTODEV.md
else
PROGRAM_PATH=
fi
If a contract exists, validate before using it:
ao autodev validate --json ${PROGRAM_PATH:+--file "$PROGRAM_PATH"}
If no contract exists and the user asked to initialize or define the loop, create one:
ao autodev init "<objective>"
Infer the objective from the user request when it is clear. Ask only when the objective cannot be discovered from repo context and inventing it would make the contract misleading.
Step 2: Repair or explain validation failures
When validation fails, inspect the missing fields and patch the program file if the user asked to create or fix the contract. Required sections:
ObjectiveMutable ScopeImmutable ScopeExperiment UnitValidation CommandsDecision PolicyEscalation RulesStop Conditions
Prefer narrow mutable scope and concrete validation commands. If the needed work crosses immutable scope, create or update a bead instead of silently widening the contract.
Step 3: Hand off to the loop
After ao autodev validate passes:
- For one lifecycle, run
$rpi "<goal>"orao rpi .... - For the repeated autonomous loop, run
$evolveorao evolve --max-cycles <n>. - If both
PROGRAM.mdandGOALS.mdexist,GOALS.mdis strategic fitness andPROGRAM.mdis the operational execution layer.
Do not mark an autonomous cycle successful only because the main tests pass. The program validation bundle and stop conditions must also be satisfied.
Examples
User: turn this postmortem/analyze/plan/pre-mortem/implement/validate loop into
a v2 command.
Agent: Explain that `ao evolve` runs the loop, then create or validate
`PROGRAM.md` with `$autodev` so the loop has explicit scope and gates.
ao autodev init "Continuously improve AgentOps skills within explicit scope."
ao autodev validate
ao evolve --max-cycles 1
Troubleshooting
| Problem | Response |
|---|---|
PROGRAM.md not found |
Run ao autodev init "<objective>" when setup is requested. |
| validation reports missing sections | Patch the missing required sections, then rerun ao autodev validate --json. |
| requested work is outside immutable scope | Stop direct edits and create a bead or ask for an explicit contract change. |
| user asks "is this evolve?" | Answer: autodev defines the loop contract; evolve runs the loop. |
Recommended Agent Skills
Expand your agent's capabilities with these related and highly-rated skills.
swarm
Spawn isolated Codex sub-agents for parallel task execution using the current runtime primitives. Triggers: "swarm", "spawn agents", "parallel work", "run in parallel", "parallel execution".
council
Multi-perspective review for Codex using the current sub-agent runtime. Triggers: "council", "get consensus", "multi-model review", "multi-perspective review", "council validate", "council brainstorm", "council research".
openai-docs
Use when the user asks how to build with OpenAI products or APIs and needs up-to-date official documentation with citations (for example: Codex, Responses API, Chat Completions, Apps SDK, Agents SDK, Realtime, model capabilities or limits); prioritize OpenAI docs MCP tools and restrict any fallback browsing to official OpenAI domains.
crank
Hands-free epic execution for Codex using wave-based sub-agents and lead-side validation. Triggers: "crank", "run epic", "execute epic", "run all tasks", "hands-free execution", "crank it".
pr-retro
Learn from PR outcomes. Analyzes accept/reject patterns and updates contribution lessons. Triggers: "pr retro", "learn from PR", "PR outcome", "why was PR rejected", "analyze PR feedback".
ratchet
Brownian Ratchet progress gates for RPI workflow. Check, record, verify. Triggers: "check gate", "verify progress", "ratchet status".
Didn't find tool you were looking for?