agentleFS
58254727-blip58254727-blip/codex-task-control-tower/skills/execution-controller/SKILL.md

execution-controller

Skill1 starsChanged 3 months agoMITView on GitHub

Use when a bounded software objective should advance end to end through planning, Skill selection, implementation, evidence tracking, and final verification.

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 4 are distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read the objective, repository state, existing evidence, frozen behavior, and system and user boundaries.

    line 14
  2. Very specific:

    Use development-planner, or apply its task contract as a conservative fallback, to create or validate the smallest sufficient graph.

    line 16
  3. Very specific:

    Use skill-router, or its route contract, for each dependency-ready task.

    line 18
  4. Distinctive:

    After each meaningful action, update the evidence ledger.

    line 22
  5. Very specific:

    Run verification-gate before reporting completion.

    line 23
  6. Very specific:

    When durable machine-readable state is useful, use codex-control-tower as an optional local evidence ledger.

    line 33
  7. Distinctive:

    After two equivalent failures, stop that path.

    line 44
  8. Distinctive:

    Do not restart the whole objective to hide a local failure.

    line 46
  9. Very specific:

    Do not mark a task complete from activity metadata or intention.

    line 51

≈ “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

Execution Controller

This is the plugin's end-to-end entry point. It advances one approved objective through the existing Codex session; it is not a background service and does not continue after the runtime stops.

Control Loop

  1. Read the objective, repository state, existing evidence, frozen behavior, and system and user boundaries.
  2. Use development-planner, or apply its task contract as a conservative fallback, to create or validate the smallest sufficient graph.
  3. Use skill-router, or its route contract, for each dependency-ready task.
  4. Execute only dependency-ready work. Keep the critical path in the current task unless another execution mechanism is both available and appropriate.
  5. Parallelize only when the runtime permits it and write scopes do not overlap.
  6. After each meaningful action, update the evidence ledger.
  7. Run verification-gate before reporting completion.

Evidence Ledger

For each task record its state, latest evidence, evidence time, write scope, attempt count, blocker, and next safe action. Evidence means an inspectable diff, command result, test, runtime log, commit, or external receipt.

Optional Local Evidence Ledger

When durable machine-readable state is useful, use codex-control-tower as an optional local evidence ledger. It can initialize a bounded task graph, record evidence and equivalent-failure keys, show dependency-ready work, enforce the two-failure stop, evaluate the verification gate, and create a sanitized handoff. The CLI does not replace judgment, execute project commands by itself, or grant authority for an operation. The active Codex session remains responsible for choosing and performing approved work.

Failure Control

  • Classify failures by root cause, not wording alone.
  • After two equivalent failures, stop that path. Make no third blind attempt.
  • Preserve files, diffs, logs, and the last known state at a safe checkpoint.
  • Do not restart the whole objective to hide a local failure.
  • Continue unrelated dependency-ready work only when its scope remains safe.

Completion And Pause

Do not mark a task complete from activity metadata or intention. If verification is unavailable, report unverified. If execution pauses, use task-handoff for a sanitized continuation record. Use task-control-tower only when multiple tasks need an evidence-backed board.

Planning and Skill routing never override system and user boundaries, install missing capabilities, or authorize credentials, production data, publishing, deployment, or destructive operations.

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.