Skip to content

dayfinggg/openai-codex-agent-skills / codex/skills/autonomous-run

autonomous-run

Verified

Carry an explicitly unattended, headless, or multi-session task through its stated finish condition. Use when the user says to continue without check-ins, work until a result, or resume long-running work after compaction. Do not infer a persistent Codex Goal from an ordinary task.

FileKindSize
SKILL.mdskill md1.3 KB
references/long-runs.mdmarkdown1.6 KB

references/long-runs.md

View on GitHub
# Long-run records and stopping conditions

Keep a compact progress file in the session scratchpad when the run may resume after compaction or in another session. Record the task's finish condition, completed and open work, current state, important decisions, check results, and what still depends on the user. Keep credentials and unnecessary personal data out of the file. Update it after meaningful work, not after every tool call.

When resuming, verify earlier claims against the current files, revision history, and live state before relying on them. A summary or worker report is a lead, not proof.

Keep working through ordinary command failures, timeouts, and discoverable missing context. Diagnose failures before changing the approach. If the same check fails repeatedly, inspect the evidence and stop retrying the same premise. Continue independent tasks while a genuinely asynchronous job runs.

Stop only when the requested finish condition is met or when a concrete dependency remains outside the authorized scope. For an unresolved decision, an approval boundary, denied access, or unavailable credential, finish all independent work, keep the exact remaining step prepared when safe, and report what input is required. An unattended run does not gain permission to publish, commit, deploy, or change shared systems unless the request explicitly covers that action.

If a tool call must continue beyond the current turn, identify the job by its returned identifier and say that the work is unfinished and what result is pending. Do not describe the task as complete until the relevant check covers the latest changes.
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall