agentleFS
bohewubohewu/agents_pipeline/skills/run-flow/SKILL.md

run-flow

Skill1 starsChanged 4 months agoMITView on GitHub

Adopt and execute the installed agents_pipeline bounded flow workflow with at most five tasks, risk-derived verification, and an optional reviewer gate. Use when the user explicitly invokes `$run-flow` or requests the normal multi-task agents_pipeline engineering flow.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 0 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use the installed orchestrator-flow definition as the authoritative workflow.

    line 8
  2. Very specific:

    Read ${CODEX_HOME:-$HOME/.codex}/agents/orchestrator-flow.toml as the authoritative workflow.

    line 10
  3. Very specific:

    If the global definition is absent, stop and ask for the global agents_pipeline Codex install.

    line 12
  4. Very specific:

    Remove only the $run-flow token and preserve the remaining request and flags as raw input.

    line 13

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Run Flow

Use the installed orchestrator-flow definition as the authoritative workflow.

  1. Read ${CODEX_HOME:-$HOME/.codex}/agents/orchestrator-flow.toml as the authoritative workflow. Never manually adopt a raw workspace role; the preflight below decides whether effective Codex config may apply workspace routing.
  2. Always query the installed global profile manager for current-workspace JSON status before adopting the workflow. A normal workspace without a profile reports global inheritance and may continue. If status cannot be verified or a configured profile's health is not ok, stop and ask the user to rerun workspace set or clear; never dispatch through an unhealthy or orphaned profile. If a configured profile's catalog_state is not current, stop before dispatch and require explicit workspace set --model-set openai or clear. If its profile_eligibility is not eligible, warn that Codex is ignoring the workspace layer and continue with global role routing. Only a healthy, eligible, current profile may route dispatched roles locally. For that eligible path, retain each role's exact saved resolved_configuration from the selected versioned model set and registered reasoning projection, including its configuration identity, version, and digest. Pass that same envelope to the formal shared resolver and trace expectations. The resolver selects effort only, so native managed normal dispatch omits a raw model; a Web session agent Job selects only the saved role model under protocols/REASONING_POLICY.md. Retired Codex model sets are not eligible for new dispatch.
  3. If the global definition is absent, stop and ask for the global agents_pipeline Codex install. Do not reconstruct the workflow from memory.
  4. Remove only the $run-flow token and preserve the remaining request and flags as raw input.
  5. Adopt the definition in the current/main agent; do not spawn the same orchestrator merely to enter the mode.
  6. Obey its task bound, delegation, verification, review, cleanup, reasoning, and final-report constraints. The formal shared resolver resolves each dispatched role's effort without changing the current/main agent. Native managed dispatch uses effective Codex role routing, normally omits a model override, and verifies the local child trace before claiming adaptive enforcement. A Web session agent Job instead selects the saved role model explicitly, applies effort only in adaptive mode, observes one Job to terminal, and follows the result and status limits in protocols/REASONING_POLICY.md; it never claims native role identity or trace enforcement. The sole native managed auto-recovery exception remains in protocols/CAPABILITY_RECOVERY.md and requires matching trace evidence. --review=max remains an exact reviewer-only effort gate: adaptive requests and verifies it on the native surface, shadow records it without applying it, and inherit conflicts. It remains deep ordinary review, never applies to a non-review role, and is unavailable on the Job surface until supported evidence exists.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.