hyq718/autoresearch
skills/autoresearch/SKILL.md
Turn a single mission markdown such as autoresearch.md into a persistent Codex research runtime with on-disk state, repeated codex exec bursts, job polling, and optional Lark progress updates. Use when the user wants long-running autonomous research that survives waiting on jobs or interrupted sessions.
$ install --global
skillsauthnpx skillsauth add hyq718/Auto-Codex autoresearchInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
Security Scan Results
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
3
Scanners Passed
9
Scanners in report
Clean
TrivyContainer and dependency vulnerability scanner
95%
Clean
SemgrepStatic code analysis for vulnerabilities
95%
Clean
mcp-scan (Snyk)Model Context Protocol security validation
95%
Skipped
Snyk (dep)Open source security scanning
50%
Skipped
Socket.devSupply chain security analysis
50%
Skipped
VirusTotalMulti-engine malware detection
50%
Skipped
CrowdStrikeAdvanced threat intelligence
50%
Skipped
OSV-ScannerOpen Source Vulnerability database check
50%
Skipped
OWASP Dep-Check
50%
Last scanned: Apr 25, 2026, 1:37 PM249.9s1 file scanned
SKILL.md
- name:
- autoresearch
- description:
- Turn a single mission markdown such as autoresearch.md into a persistent Codex research runtime with on-disk state, repeated codex exec bursts, job polling, and optional Lark progress updates. Use when the user wants long-running autonomous research that survives waiting on jobs or interrupted sessions.
autoresearch
Use this skill when the user wants Codex to keep pushing a research or engineering task from one mission markdown instead of relying on one long chat session.
The repository now also includes a conversation-mode adapter that exposes the runtime as /autoresearch-style commands on top of the same persisted state.
Workflow
- Treat the mission markdown as the single source of truth.
- Bootstrap a runtime with
scripts/autoresearch.py init. - Start or resume the supervisor with
scripts/autoresearch.py startor run it in the background withdaemon-start. - Keep progress in
state.json,plan.json,events.jsonl,jobs/, andnotes/latest_summary.md. - Persist out-of-band user feedback with
add-input/list-inputs/ack-input. - Let the supervisor inject pending inputs into the worker prompt and acknowledge consumed ones.
- Append meaningful updates to the mission's Lark document when a doc URL is available.
- Let the supervisor poll the same Lark document for new user-visible content and treat it as input.
- Use the Slurm helper commands when you need a small standardized
sbatch/squeuepath. - Prefer the
mode-*commands when the user wants a chat-visible control surface rather than raw runtime plumbing.
Commands
python3 scripts/autoresearch.py init /path/to/autoresearch.md --runtime-dir /path/to/runtime
python3 scripts/autoresearch.py mode-start /path/to/runtime --mission /path/to/autoresearch.md
python3 scripts/autoresearch.py mode-status /path/to/runtime
python3 scripts/autoresearch.py mode-sync /path/to/runtime
python3 scripts/autoresearch.py mode-update /path/to/runtime --message "New suggestion"
python3 scripts/autoresearch.py mode-plan /path/to/runtime
python3 scripts/autoresearch.py mode-jobs /path/to/runtime
python3 scripts/autoresearch.py mode-pause /path/to/runtime
python3 scripts/autoresearch.py mode-resume /path/to/runtime
python3 scripts/autoresearch.py mode-stop /path/to/runtime --reason "manual stop"
python3 scripts/autoresearch.py start /path/to/runtime --search
python3 scripts/autoresearch.py daemon-start /path/to/runtime --search
python3 scripts/autoresearch.py status /path/to/runtime --json
python3 scripts/autoresearch.py daemon-status /path/to/runtime --json
python3 scripts/autoresearch.py add-input /path/to/runtime --message "New suggestion"
python3 scripts/autoresearch.py list-inputs /path/to/runtime --pending-only --json
python3 scripts/autoresearch.py ack-input /path/to/runtime <input-id> --resolution "Queued"
python3 scripts/autoresearch.py submit-job /path/to/runtime ./train.sbatch --notes "smoke test"
python3 scripts/autoresearch.py sync-jobs /path/to/runtime --json
python3 scripts/autoresearch.py list-jobs /path/to/runtime --json
python3 scripts/autoresearch.py stop /path/to/runtime --reason "manual stop"
python3 scripts/autoresearch.py daemon-stop /path/to/runtime --reason "manual stop"
Guidance
- Prefer short, resumable bursts. The supervisor is responsible for waking Codex again.
- If the mission already contains a Feishu/Lark doc URL, let the runtime use it as the default progress target.
- Pending inputs can come from both local runtime commands and Feishu polling.
mode-updateshould be treated as the chat bridge into the same persisted input queue used by the supervisor.mode-statusandmode-syncare the preferred user-facing views when you want a stable, readable progress report in the active conversation.- When the user asks for packaging or publishing, treat this repository as a reusable
skill + scriptsproject rather than a single prompt file.
Related Skills
hyq718/auto-codex
development
VerifiedTrustedCommunity
Run or steer the Auto-Codex persistent autoresearch runtime from one mission markdown. Use when the user wants a `/autoresearch`-style workflow in Codex, wants to bootstrap a runtime from `autoresearch.md`, inspect or sync progress, inject new chat directions, pause or resume work, or drive the same long-running runtime that also reports to Feishu and tracks jobs on disk.
1SKILL.mdUpdated Apr 25, 2026
hyq718/auto-codex
development
VerifiedTrustedCommunity
Run or steer the Auto-Codex persistent autoresearch runtime from one mission markdown. Use when the user wants a `/autoresearch`-style workflow in Codex, wants to bootstrap a runtime from `autoresearch.md`, inspect or sync progress, inject new chat directions, pause or resume work, or drive the same long-running runtime that also reports to Feishu and tracks jobs on disk.
1SKILL.mdUpdated Apr 25, 2026
openclaw/openclaw-secret-scanning-maintainer
development
VerifiedTrustedCommunity
Maintainer-only workflow for handling GitHub Secret Scanning alerts on OpenClaw. Use when Codex needs to triage, redact, clean up, and resolve secret leakage found in issue comments, issue bodies, PR comments, or other GitHub content.
357,764SKILL.mdUpdated Apr 15, 2026
openclaw/openclaw-release-maintainer
development
VerifiedTrustedCommunity
Maintainer workflow for OpenClaw releases, prereleases, changelog release notes, and publish validation. Use when Codex needs to prepare or verify stable or beta release steps, align version naming, assemble release notes, check release auth requirements, or validate publish-time commands and artifacts.
357,764SKILL.mdUpdated Apr 10, 2026