SkillRepo
Sign InSign Up
Catalog
yeachan-heo/autoresearch

yeachan-heo

autoresearch

Stateful single-mission improvement loop with strict evaluator contract, markdown decision logs, and max-runtime stop behavior

v1.0LATEST
NewUpdated Sep 9, 2026

<Use_When>

  • You already have a mission and evaluator from /deep-interview --autoresearch
  • You want persistent single-mission improvement with strict evaluation
  • You need durable experiment logs under .omc/autoresearch/
  • You want a supported path for periodic reruns via Claude Code native cron </Use_When>

<Do_Not_Use_When>

  • You need evaluator generation at runtime — use /deep-interview --autoresearch first
  • You need multiple missions orchestrated together — v1 forbids that
  • You want the deprecated omc autoresearch CLI flow — it is no longer authoritative </Do_Not_Use_When>

<Required_Artifacts> Canonical persistent storage lives under .omc/autoresearch/<mission-slug>/ and/or .omc/logs/autoresearch/<run-id>/.

Minimum required artifacts:

  • mission spec
  • evaluator script or command reference
  • per-iteration evaluation JSON
  • markdown decision logs

Recommended canonical shape:

.omc/autoresearch/<mission-slug>/
  mission.md
  evaluator.json
  runs/<run-id>/
    evaluations/
      iteration-0001.json
      iteration-0002.json
    decision-log.md

Reuse existing runtime artifacts when available rather than duplicating them unnecessarily. </Required_Artifacts>

<Cron_Integration> Claude Code native cron is a supported integration point for periodic mission enhancement. In v1, prefer documenting/configuring cron inputs over building a large scheduler UI.

If cron is used:

  • keep one mission per scheduled job
  • preserve the same mission/evaluator contract
  • append new run artifacts rather than overwriting prior experiments </Cron_Integration>

<Execution_Policy>

  • Do not hand execution back to omc autoresearch
  • Do not create multi-mission orchestration
  • Prefer reusing src/autoresearch/* runtime/schema helpers where they already match the stricter contract
  • Keep logs useful to humans, not only machines </Execution_Policy>
Files1
1 files · 1.0 KB

Select a file to preview

Overall Score

82/100

Grade

B

Good

Grades are signals, not a certification. Always review a skill yourself before use.

Safety

85

Quality

83

Clarity

82

Completeness

76

Summary

Autoresearch is a stateful skill for bounded, evaluator-driven iterative improvement of a single mission. It maintains durable artifacts (evaluation JSON, markdown decision logs) in `.omc/autoresearch/` directories, orchestrates experiment-evaluate-log cycles with explicit stop conditions (max-runtime, user cancellation), and integrates with Claude Code native cron for periodic reruns. The skill enforces a strict contract: single mission per run, evaluator must return structured JSON with boolean `pass` field, and non-passing iterations continue until terminal conditions are met.

Detected Capabilities

file readfile writestructured JSON generationdirectory structure managementmission state trackingevaluator invocationmarkdown log appendingtimestamp recording

Trigger Keywords

Phrases that agents use to match this skill to user intent.

autoresearch iteration loopbounded improvement missionevaluation-driven experimentpersistent experiment logsmission enhancement rerunsmax-runtime stop control

Risk Signals

INFO

No file deletion or destructive operations

Workflow section
INFO

File writes scoped to .omc/autoresearch/ and .omc/logs/autoresearch/ project directories

Required_Artifacts section
INFO

Evaluator invocation referenced but not executed directly by skill — delegated to existing evaluator script/command

Workflow step 3
INFO

Max-runtime stop condition is explicit and bounded — prevents infinite loops

Workflow step 4 and Execution_Policy

Use Cases

  • Run bounded iterative improvements on a single mission with persistent experiment logs
  • Integrate mission enhancement into Claude Code native cron schedules for periodic reruns
  • Maintain durable evaluation records and human-readable decision logs for experiment tracking
  • Stop only at explicit max-runtime ceilings or user-triggered terminal conditions
  • Reuse mission/evaluator setup from deep-interview --autoresearch in a strictly controlled loop

Quality Notes

  • Excellent contract specification: clearly defines single-mission boundary, required evaluator JSON schema (boolean pass, optional numeric score), and stop conditions
  • Well-structured artifact canonical shape with clear directory hierarchy and run isolation under .omc/autoresearch/<mission-slug>/runs/<run-id>/
  • Explicitly forbids multi-mission orchestration and references deprecated CLI flow — reduces confusion and scope creep
  • Workflow steps are logically sequenced with clear per-iteration semantics (experiment → evaluate → persist → log → continue or stop)
  • Max-runtime as primary strict stop hook is well-documented and prevents runaway execution
  • Markdown decision logs + machine-readable JSON separation supports both human review and automated analysis
  • Cron integration guidance is practical and preserves mission/evaluator contract across periodic reruns
  • Clear deprecation of omc autoresearch CLI in favor of this skill-driven flow reduces ambiguity
Model: claude-haiku-4-5-20251001Analyzed: Sep 9, 2026

Reviews

Add this skill to your library to leave a review.

No reviews yet

Be the first to share your experience.

Use yeachan-heo/autoresearch in your dev environment

Start a Developer trial

Command Palette

Search for a command to run...