Run any Skill in Manus with one click

propose-harness-improvement

Stars2

Forks0

UpdatedMay 30, 2026 at 14:54

Use this skill whenever the agent makes a mistake, the user observes an avoidable failure, a pattern recurs, or someone says "the agent keeps doing X". Files an "Engineer the Harness" entry — Mitchell Hashimoto's discipline: every failure becomes a permanent prevention mechanism. Always invoke this instead of just fixing the immediate symptom.

Installation

Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.

Run Skill in Manus

Source

tuanle96

tuanle96/zca-rust

View GitHub Repository View Creator Repositories

Download

Run Skill in Manus

Related occupationsSOC

Based on SOC occupation classification

Software DevelopersComputer and Mathematical Occupations·SOC 15-1252

File Explorer

4 files

SKILL.md

readonly

name	propose-harness-improvement
description	Use this skill whenever the agent makes a mistake, the user observes an avoidable failure, a pattern recurs, or someone says "the agent keeps doing X". Files an "Engineer the Harness" entry — Mitchell Hashimoto's discipline: every failure becomes a permanent prevention mechanism. Always invoke this instead of just fixing the immediate symptom.

Steps

Triage. Ask: "What just went wrong? What was the agent's intended behavior? What's the symptom?"
Gather deterministic signals. Run the runtime-agnostic bundle when local telemetry/bypass logs exist: node .harness/scripts/improvement-bundle.mjs. Use its taxonomy_classification, suggested_records, and nextSteps as evidence, not as a substitute for judgment.
Classify. Read .harness/failures/taxonomy.json and pick exactly one primary class: context-miss, false-done, architecture-drift, test-gap, doc-drift, tool-misuse, permission-gap, runtime-gap, eval-gap, cost-spike, or model-behavior.
Write a failure record with .harness/scripts/record-failure.mjs. Include evidence paths, prevention target, proposed prevention, and verification command when the prevention is already applied or verified. If the prevention target differs from the taxonomy class preferredPrevention, include a concrete prevention justification. Hand-write JSON only if the helper is unavailable. To promote an existing record after applying the prevention, use .harness/scripts/record-failure.mjs --update=<id> --status=applied|verified so the helper preserves the original failure fields and reruns the checker.
Append a human summary to .harness/docs/agent-failures.md with: date, symptom, failure class, prevention target, file modified.
Apply the fix in the right place. NEVER paper over with a CLAUDE.md "be careful" sentence unless rule (a) applies — and even then, only as a pointer to a longer doc.
Update PROGRESS. Append harness-improvement: <slug> to .harness/PROGRESS.md.

Output contract

### Failure: <one-line summary>
### Classification: <taxonomy class>
### Prevention target: <target>
### Fix applied at: <file:line>
### Failure record: .harness/failures/records/<id>.json
### .harness/docs/agent-failures.md entry: §<n>

Anti-patterns (block on these)

Don't add a vague "be careful with X" sentence to CLAUDE.md.
Don't add a rule whose enforcement is also LLM-based.
Don't use this skill to log unrelated cleanup ideas — those go in .harness/docs/tech-debt-tracker.md.

More from this repository

same repository

add-adr

tuanle96/zca-rust

Use this skill whenever a decision is made about architecture, dependencies, frameworks, naming conventions, or layer order. Creates a numbered ADR (Architecture Decision Record) in `.harness/docs/adr/` in the canonical Nygard format. Always invoke this before changing layer order, adding a layer, swapping a major dependency, or introducing a new external service.

2026-05-302

add-feature

tuanle96/zca-rust

Use this skill whenever the user asks to add, implement, or build a new feature, capability, endpoint, page, command, or anything user-visible. Enforces the Anthropic two-fold harness pattern — read .harness/feature_list.json, pick exactly one feature, implement incrementally, run the structural test on every save, and never declare "done" without updating the JSON. Always invoke this skill instead of writing new feature code freehand.

2026-05-302

debug-flow

tuanle96/zca-rust

Use this skill whenever the user reports a bug, unexpected output, or "this doesn't work". Runs the dev server, drives the failing flow via Playwright MCP if installed (else captures stdout/stderr), and produces a minimal repro before any fix. Mirrors the OpenAI Chrome-DevTools-Protocol-into-runtime pattern at solo scale — verify the failure before you propose a fix.

2026-05-302

deliver-html

tuanle96/zca-rust

Use this skill whenever the user asks to analyze, audit, review, summarize, produce a report, write a plan, make a proposal, draft a decision doc, list "next actions", or any other task that produces a DOCUMENT a HUMAN reads-and-acts-on. Outputs a self-contained <slug>.html at repo root using the shared dark-theme CSS. Why HTML and not Markdown: golden principle #9 — Markdown is great for files an agent reads-and-edits (CLAUDE.md, SKILL.md, ADRs), but a HUMAN reading a 700-line MD deliverable will scroll, miss the conclusion, and ask the agent to clarify — burning turns and tokens. HTML deliverable is read once, decided once. Do NOT use this skill for files the agent itself reads (those stay MD), for stdout output from /review-this-pr or /garbage-collection (pass-through MD), or for short summaries under ~30 lines (overhead not worth it).

2026-05-302

doc-drift-scan

tuanle96/zca-rust

Use this skill weekly, before releases, or when the user mentions "stale docs", "doc drift", "docs are wrong", or "the README is out of date". Cross-checks every code path, file path, and command referenced in `.harness/docs/` and `CLAUDE.md` against the current repo state and produces a list of stale references — the doc-gardening agent pattern.

2026-05-302

eval-runner

tuanle96/zca-rust

Use this skill whenever a skill, subagent, or hook is changed, before merging to main, or when the user mentions "eval", "regression test for the harness", or "is the harness still working". This skill is a thin wrapper — it runs a single shell command (`npm run harness:eval` or `python -m harness.eval_runner`) and summarizes the JSONL output. Do not implement the eval logic yourself; the runner is already deterministic.

2026-05-302

name	propose-harness-improvement
description	Use this skill whenever the agent makes a mistake, the user observes an avoidable failure, a pattern recurs, or someone says "the agent keeps doing X". Files an "Engineer the Harness" entry — Mitchell Hashimoto's discipline: every failure becomes a permanent prevention mechanism. Always invoke this instead of just fixing the immediate symptom.

Steps

Triage. Ask: "What just went wrong? What was the agent's intended behavior? What's the symptom?"
Gather deterministic signals. Run the runtime-agnostic bundle when local telemetry/bypass logs exist: node .harness/scripts/improvement-bundle.mjs. Use its taxonomy_classification, suggested_records, and nextSteps as evidence, not as a substitute for judgment.
Classify. Read .harness/failures/taxonomy.json and pick exactly one primary class: context-miss, false-done, architecture-drift, test-gap, doc-drift, tool-misuse, permission-gap, runtime-gap, eval-gap, cost-spike, or model-behavior.
Write a failure record with .harness/scripts/record-failure.mjs. Include evidence paths, prevention target, proposed prevention, and verification command when the prevention is already applied or verified. If the prevention target differs from the taxonomy class preferredPrevention, include a concrete prevention justification. Hand-write JSON only if the helper is unavailable. To promote an existing record after applying the prevention, use .harness/scripts/record-failure.mjs --update=<id> --status=applied|verified so the helper preserves the original failure fields and reruns the checker.
Append a human summary to .harness/docs/agent-failures.md with: date, symptom, failure class, prevention target, file modified.
Apply the fix in the right place. NEVER paper over with a CLAUDE.md "be careful" sentence unless rule (a) applies — and even then, only as a pointer to a longer doc.
Update PROGRESS. Append harness-improvement: <slug> to .harness/PROGRESS.md.

Output contract

### Failure: <one-line summary>
### Classification: <taxonomy class>
### Prevention target: <target>
### Fix applied at: <file:line>
### Failure record: .harness/failures/records/<id>.json
### .harness/docs/agent-failures.md entry: §<n>

Anti-patterns (block on these)

Don't add a vague "be careful with X" sentence to CLAUDE.md.
Don't add a rule whose enforcement is also LLM-based.
Don't use this skill to log unrelated cleanup ideas — those go in .harness/docs/tech-debt-tracker.md.