Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add nnabuuu/harness-engineering-toolkit --skill harness-retrogit clone --depth 1 https://github.com/nnabuuu/harness-engineering-toolkitWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/nnabuuu/harness-engineering-toolkit/harness-retro)<a href="https://agentmods.dev/skills/nnabuuu/harness-engineering-toolkit/harness-retro"><img src="https://agentmods.dev/badge/skills/nnabuuu/harness-engineering-toolkit/harness-retro/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/nnabuuu/harness-engineering-toolkit/harness-retro"><img src="https://agentmods.dev/badge/skills/nnabuuu/harness-engineering-toolkit/harness-retro.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00130 | $0.03424 |
| Opus 5 | $0.00065 | $0.01712 |
| Sonnet 5 | $0.00026 | $0.00685 |
| Haiku 4.5 | $0.00013 | $0.00342 |
Grade A, and why
harness-retro scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 8d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 324 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Harness Retrospective
Analyze completed harness runs to extract patterns and generate concrete improvements to prompts, rubrics, and exit conditions. This is L4 — the evaluation loop applied to the harness system itself.
Self-Review Checklist (read BEFORE and AFTER every phase)
Discovery Guards
- Scanned ALL task directories in
.harness-workspace/, not just first match - Verified each task has enough data to analyze (at minimum: state.json or progress.md)
- Confirmed scope with user before starting analysis
Analysis Guards
- Read ALL eval reports and changelogs for selected tasks
- Did not skip any dimension — marked "insufficient data" if data missing
- All findings cite specific evidence (iteration numbers, scores, file paths)
- Cross-task analysis (D6) only when 2+ tasks selected
Report Guards
- Improvement suggestions include concrete diffs, not just descriptions
- Suggestions reference specific dimension names from EVAL_CRITERIA.md
- Report saved to RETRO.md in the task directory (single-task) or
.harness-workspace/RETRO-{date}.md(multi-task)
Archive Guards
- Archive only after retro report is complete (Entry A) OR after user acknowledges no RETRO.md (Entry B)
- User explicitly confirmed before any file operations
- Entire task directory moved to
_archive/, not left split across two locations - DAGU symlink updated to point to new
_archive/path (execution history preserved)
Entry Point Detection
Determine which entry point the user needs before starting any phase.
- Entry A — Full Retro: Phases 1 → 2 → 3 → 4 (archive optional at end). The complete retrospective flow.
- Entry B — Standalone Archive: Skip directly to Phase 4A. Archive completed/failed tasks without running analysis.
Detection Rules
| User says | Entry |
|---|---|
| "archive harness", "archive task", "clean up workspace", "归档harness", "归档" | Entry B |
| "retro", "retrospective", "review harness", "analyze harness", "复盘", "回顾", "总结harness" | Entry A |
| Unclear or ambiguous | Ask: "Do you want a full retrospective (analysis + improvements), or just archive completed tasks?" |
What ships with it
3 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 8d ago First seen · 324 lines · 130 tokens per session scan A 5c130d3ccddc
harness-retro is a skill published in the GitHub repository nnabuuu/harness-engineering-toolkit (5 stars, last pushed 1mo ago), licensed MIT. It adds 130 tokens to every session and 3,424 once invoked, about $0.0006 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
opencli-autofix
Automatically fix broken OpenCLI adapters when commands fail. Load this skill when an opencli command fails — it guides you through collecting a trace artifact, patching the adapter, retrying, and filing an upstream GitHub issue after a verified fix. Works with any AI agent.
RootCauseAnalysis
Structured incident investigation using Five Whys, Fishbone, blameless Postmortem, Fault Tree, Kepner-Tregoe, and FMEA — traces failures to systemic root causes rather than blaming humans. USE WHEN root cause, RCA, 5 whys, fishbone, postmortem, incident analysis, fault tree, why does this keep failing, blameless…
Vitals
Read-only macOS performance inspection: a deterministic CLI gathers the numbers, Interpretation.md turns them into a diagnosis instead of a data dump.
omh-failure-signal-audit
This is a Hermes-native failure-signal-audit workflow skill.
report-issue-local
File a bug or feature request against this MCP server's own repo. Use for server-specific issues — tool logic, service integrations, config problems, or domain bugs that aren't caused by the framework.
thinking-five-whys-plus
When a fault is localized and the proximate cause is known but the systemic root is not, chain evidence-linked whys with a counterfactual stop and a countermeasure.