Skip to content
Back to skills

He Compound

ASecurity

Run a bounded Harness Engineering lifecycle across multiple stages. Use when the user wants coordinated brainstorm, spec, plan, work, review, and fix flow rather than one isolated stage.

  • 9 stars
  • 0 votes
  • 0 copies
  • 2 views
  • Added June 1, 2026
ai-agentsrustdebugging

Security analysis

A100/100

Scanned June 1, 2026

npx -y skills add jscraik/Agent-Skills --skill he-compound --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of He Compound?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for He Compound
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/jscraik-he-compound/badge)](https://www.skillsdirectory.com/skills/jscraik-he-compound)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: he-compound
description: Run a bounded Harness Engineering lifecycle across multiple stages. Use when the user wants coordinated brainstorm, spec, plan, work, review, and fix flow rather than one isolated stage.
metadata:
  skill-type: team_automation
---

# Harness Engineering Compound

Progressive-disclosure entrypoint for stage orchestration and durable learning capture in Harness Engineering.

## Philosophy

- Route to the correct stage before acting.
- Capture verified outcomes as reusable team knowledge.
- Keep orchestration, evidence gathering, and durable learning capture explicit instead of collapsing them into generic advice.

## When to use

- Route/resume work from the correct Harness Engineering stage.
- Capture a verified fix in `docs/solutions/` and avoid duplicate docs by refreshing existing artifacts when overlap is high.
- Decide whether a request needs lifecycle orchestration, solved-problem capture, or a resume-from-stage handoff before downstream work starts.

## Inputs

- Request, artifacts, repo context, and linked Linear issues.

## Outputs

- `schema_version: 1` when structured; result, validation, blockers, and next Harness Engineering action.

## Procedure

1. Select lifecycle mode using artifact-first evidence.
2. For lifecycle routing, continue from the earliest incomplete or untrusted stage.
3. If the user intent is ambiguous between stage orchestration and solved-problem capture, ask one direct question before continuing.
4. For learning capture, preserve `full` as the default and treat `compact-safe` as explicit opt-in.
5. For learning capture, validate solved evidence, gather supporting inputs, and write exactly one durable solution artifact.
6. If past agent sessions are needed as evidence, route broad session inventory/extraction to `skill-refactor`; if the result is a reusable workflow, route skill creation to `skillify`.
7. If helpers are used during learning capture, they return text only; the orchestrator writes the final artifact.
8. If overlap with an existing solution is high, refresh the existing doc instead of creating a duplicate.
9. Recommend `he-compound-refresh` only when adjacent stale or overlapping docs need selective follow-up beyond the current artifact.

## Validation

- Confirm mode selection matches available evidence.
- Confirm stage recommendation names the exact next Harness Engineering command or stage.
- For learning capture, verify solved status before writing docs, keep the one-file-write contract explicit, and ensure duplicate-avoidance logic is explicit.
- Any refresh recommendation must be narrow and evidence-backed rather than a blanket cleanup suggestion.
- Fail fast: stop at first failed gate and do not continue.

## Constraints

- Redact secrets, credentials, tokens, and sensitive data by default.
- Do not document unverified fixes as durable solutions.
- Do not treat `he-compound` as a substitute for implementation, debugging, review, or refresh stages that still need to happen.
- Do not turn generic session-history search into a Harness Engineering stage; consume `skill-refactor` or `skillify` outputs when session evidence is relevant.
- Do not let helper agents write intermediate files during learning capture.
- Do not recommend deleting or ignoring protected process artifacts just to simplify the handoff.
- Apply the context-disposition policy: move important still-valid context to references and index it when meaningful; intentionally discard stale, duplicated, unsafe, superseded, or low-signal text.

## Anti-patterns

- Guessing stage progression without checking existing artifacts.
- Recording unresolved incidents as solved knowledge.
- Creating a duplicate solution doc when a high-overlap artifact should be refreshed.
- Returning broad "use Harness Engineering" guidance without naming the exact mode and next stage.

## Full Context

- Assets: [icon-small.png](./assets/icon-small.png), [icon-large.png](./assets/icon-large.png)

## Examples

- "Can you inspect this feature's artifacts, resume from the right Harness Engineering stage, and tell me the next exact action?"
- "We fixed the same CI failure twice. Capture the solved workflow once and refresh the existing solution if it overlaps."
- "Help me coordinate brainstorm, spec, plan, work, and review for this Linear issue without skipping the earliest weak stage."

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…