minhtran3124/agent-harness/compound
- community
Checking that the SKILL.md file is still in the repository.
Preserve durable bugs, patterns, decisions, and failed approaches as discoverable `docs/solutions/` records. Use only after a session produced a reusable learning or a proposed harness ratchet.
Install
npx vskill@latest install minhtran3124/agent-harness/compoundSKILL.md
Compound — durable session knowledge
Use when a non-obvious bug, reusable pattern, considered decision, useful failure, or Harness-Delta: backlog signal is worth preserving. Do not emit a document merely because a session ended.
Extract and select
You hold the session, so run the context-analyzer, solution-extractor, and decision-extractor passes yourself, using the prompt files in subagents/ as their output schemas. Then run the related-docs pass with the context analyzer's module/tags. Dispatch a subagent only for a docs/solutions/ tree too large to screen from INDEX.md, and paste in its inputs — a fresh context cannot see this session.
Emit only complete tracks:
| Track | Required content | |
|---|---|---|
| bug | problem, root cause, fix | |
| knowledge | pattern, how to use | |
| decision | context, options, rationale | |
| failure | symptom, wrong approach, why it failed, correct approach |
Skip incomplete or [none] tracks. Preserve a relevant active spec’s alternatives as decision inputs. A spec number labelled prediction → measured (a prediction kept beside its measured <YYYY-MM-DD> <source> value) is a learning source: the gap between the two is evidence about how the team estimates.
Write canonical records
Read the selected templates/ file before writing. Use the context category/slug under docs/solutions/<category>/; on a high-overlap collision update the existing record, otherwise allocate the next numeric suffix. Multiple decisions use one -decisions consolidated template.
For a critical context, append an entry per emitted track to critical-patterns.md using its templates. For a failure with proposed: guardrail, append one open row to docs/harness-experimental/improvement-backlog.md; do not duplicate an existing: guardrail.
Run the authority at the end of every run, even when nothing was written:
Do not manually scan/sort/render INDEX.md. It excludes INDEX/critical patterns, preserves exact frontmatter values, and offers --check/--dry-run for review.
Finally verify knowledge-base discoverability in CLAUDE.md and rules. If missing, propose—not silently add—the documented knowledge-base section. Report tracks emitted, collisions, index result, critical promotion, and any proposed ratchet.
Security
- Security scanPassed, 95 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/compound/SKILL.md
Popularity
- Repository stars
- 7
- 7-day trend score
- 10
Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/minhtran3124/agent-harness/compound)