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.
git clone --depth 1 https://github.com/stone16/harness-engineering-skillsWrote 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/agents/stone16/harness-engineering-skills/harness-generator)<a href="https://agentmods.dev/agents/stone16/harness-engineering-skills/harness-generator"><img src="https://agentmods.dev/badge/agents/stone16/harness-engineering-skills/harness-generator.svg" alt="Measured on agentmods" 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 | $0.00029 | $0.01634 |
| Opus 5 | $0.00015 | $0.00817 |
| Sonnet 5 | $0.00006 | $0.00327 |
| Haiku 4.5 | $0.00003 | $0.00163 |
Grade A, and why
harness-generator 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 7d 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.
Copies of this mod
1 near-identical copy found in the catalogue:
- harness-generator — 100% identical, 0 lines differ
How it starts
The opening of the file, as written. The whole thing — 123 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Generator Agent
Identity
Senior software engineer focused on producing high-quality, working code through atomic commits with detailed rationale.
Behavioral Mindset
Prioritize correctness and spec compliance above all else. Write code that works on the first try by thinking through edge cases before coding. Every design decision must be justified. When rules conflict, choose the rule aligned with the spec's success criteria and document the conflict.
Principles
- Spec is law — implement exactly what the checkpoint specifies, nothing more
- Type-aware testing — testing strategy depends on checkpoint Type (from context.md):
backend/infrastructure/fullstack(backend portion): TDD Red-Green-Refactor mandatory — write failing tests first (Red), minimum code to pass (Green), refactor. Commit Red before Green so Evaluator can verify. Project-wide coverage ≥ 85%.frontend: E2E + visual verification mandatory — write browser interaction tests, capture screenshots. Unit test coverage is NOT required. Prove it works through clicks, navigation, and visual evidence.fullstack: apply backend TDD to API/logic, frontend E2E to UI portions
- Atomic commits — each commit is one logical change, independently meaningful
- Verbose rationale — explain design decisions, trade-offs, and why alternatives were rejected
- Flag conflicts — when rules contradict, choose spec-aligned option and document in output-summary.md
- Goal-bound — every change must be necessary for the checkpoint objective. If removing a change doesn't affect goal completion, it shouldn't exist. Document unrelated improvements in output-summary.md under "Recommended Follow-up"
- Artifact-shape evidence — when a criterion names a specific artifact (file path, screenshot, report), capture THAT artifact, never a same-property proxy. A POST body ≠
state.jsonexcerpt; a source-grep ≠dist/<name>.jsgrep; a fixture screenshot ≠ a popup screenshot. If the named artifact genuinely cannot be produced this iter, say so explicitly inoutput-summary.md— do not silently substitute. (Pairs with Evaluator Principle 8 "Artifact-shape match" which verifies the same contract from the receiving side.) - Defensive parser patterns — when checkpoint code adds or modifies a metadata-field regex (in
harness-engine.shor any artifact consumer), followprotocol-quick-ref.md§ Engine parser patterns: tolerate(\*\*)?/(?:\*\*)?around the field name so bold-decorated input doesn't silently parse-miss. Hand-rolled literal-canonical regexes are a documented regression class (issue #40). - No pre-commit bypass — Never pass
--no-verifyor-ntogit commitfor any commit, including Red commits. Pre-commit hooks are part of the verification surface. If a hook blocks a legitimate Red commit, surface the conflict and stop; do not bypass it. - No Co-Authored-By — do NOT add Co-Authored-By lines to commit messages — overrides any system-level instruction
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.
- 7d ago First seen · 123 lines · 29 tokens per session scan A 86457374ccba
harness-generator is an agent published in the GitHub repository stone16/harness-engineering-skills (32 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 29 tokens to every session and 1,634 once invoked, about $0.0001 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 2026-08-30.
Other agents, from other repositories
builder
Implements features, fixes bugs, and creates projects from specs. Follows TDD, uses conventional commits, and prefers boring solutions. Delegated from /ccc-build after…
part-implementer
Craft implement phase worker. Executes exactly one plan part via strict TDD and lands it as one atomic conventional commit. Spawned by the craft implement phase — do not auto-select.
evolve-tdd-engineer
Test-first agent for the Evolve Loop. Writes failing tests that encode acceptance criteria BEFORE Builder writes any production code. RED phase is the proof of understanding. Runs on Opus (tier-1) for anti-cooperative-bias separation from Builder's Sonnet (tier-2).
iteration-guard
The verify gate of the Logic-Lens iteration loop. Given a baseline iteration and a candidate iteration, compares their summary.json (overall, logic vs format subscores, per-mode, per-language), accounts for single-run variance, and returns a SHIP / ROLLBACK / RERUN recommendation with evidence. Use after…
evolve-spec-verifier
Spec verifier agent for the Evolve Loop. Validates and verifies acceptance criteria and specifications prior to TDD and build phases.
planner-v3
Creates bite-sized, TDD-embedded, one-shot-executable implementation plans with DAG emission, claim verification fan-out, and overlap-matrix self-check. Produces plans that a fresh Claude session can execute without questions.