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 BCGen/skills --skill harness-syncgit clone --depth 1 https://github.com/BCGen/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/skills/bcgen/skills/harness-sync)<a href="https://agentmods.dev/skills/bcgen/skills/harness-sync"><img src="https://agentmods.dev/badge/skills/bcgen/skills/harness-sync/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/bcgen/skills/harness-sync"><img src="https://agentmods.dev/badge/skills/bcgen/skills/harness-sync.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 | $0.00083 | $0.00961 |
| Opus 5.5 | $0.00033 | $0.00384 |
| Sonnet 5.5 | $0.00017 | $0.00192 |
| Haiku 4.5 | $0.00008 | $0.00096 |
Grade A, and why
harness-sync 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 2mo 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 — 88 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Harness Sync
Set up — and keep synced — the files agents read ("instruction surface") plus the harness loop. Tables and templates live in references/playbook.md — read it first. Never touch execution settings (settings.json, hooks, MCP, CI).
Step 1 — Detect
Check for agent artifacts (.claude/, CLAUDE.md, .cursor/, AGENTS.md)
and coexisting frameworks (.specify/, .agent-os/, .kiro/steering/) per
the playbook. Zero or multiple agents → ask one question with a recommended
answer (AGENTS.md is the portable default). Never guess.
Step 2 — Inventory
List what already exists: entry file, harness block, interop glue. The plan
covers the missing pieces AND the harness block if it drifted from the
current template — re-running on an up-to-date project must be a no-op. The
.ai/ loop directories are retro's to create; harness-sync neither creates
nor reconciles them.
Step 3 — Discover (new entry file only)
If an entry file must be created, gather verified facts from the repo:
build/test/lint commands read from manifests (package.json, Makefile,
CI config), plus layout facts worth stating. Never write a command you did
not verify in a file. Where the agent has a native bootstrap (e.g. /init),
prefer it for discovery, then trim to the content policy: no generic
rules, no personas, no best-practice boilerplate — verified facts and the
harness block only.
Step 4 — Present the plan, then get approval
Print the full plan in the conversation — the complete content of every file to be created and a diff for every file to be modified — then ask for approval with the platform's option-prompt tool when it has one (Claude Code: AskUserQuestion) — choices like approve-as-planned / adjust-first; plain question only when no such tool exists. Never ask the user to approve content they have not seen. The plan covers:
- Entry file (create, append the managed harness block to an existing
one, or update a drifted block in place — never edit outside your
harness:begin/endmarkers). - Interop glue when applicable (e.g.
CLAUDE.mdstarting with@AGENTS.md). - No rules directory — git cannot track an empty directory; rule-writing creates it with the first rule.
- No
.ai/loop directories — retro creates and maintains.ai/learnings/and.ai/backlog/on its first write; harness-sync only names them in the harness block. - Coexisting framework detected → one reference line instead of duplicate content, per the playbook.
What ships with it
1 file 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.
- 2mo ago First seen · 88 lines · 83 tokens per session scan A ad048c84220b
harness-sync is a skill published in the GitHub repository BCGen/skills (2 stars, last pushed 3mo ago), licensed MIT. It adds 83 tokens to every session and 961 once invoked, about $0.0003 per session on Opus 5.5. 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
memory-system
A guide to giving an AI agent persistent, user-specific memory across conversations. It covers what to remember, how to store it, how to update or compress it, and when to use memory instead of a shared knowledge base.
eval-dataset-design
A guide to creating reliable test sets for AI agents. Each evaluation task includes the request, the simulated user's behaviour, a reset starting state, and a success check that can be independently verified.
writing
A writing guide for turning verified facts and calculations into finished text for a specific audience. It follows the requested language, structure, and length.
pkf-writing
Author, update, convert, validate, persist, and render PKF (.pkf) content through message-inline, Library Asset, or Memory Page carriers. Use for structured reports, knowledge pages, delivery cards, or any .pkf body. NOT for recall, memory-graph placement, or curation decisions — those belong to the memory skill.
food-order
Reorder previous Foodora orders, preview cart contents, and track delivery ETA/status with ordercli. Use when the user wants to reorder food, check delivery status, or browse recent Foodora order history. Never confirm an order without explicit user approval.
memory-import
Bulk-import an external knowledge base into Prismer memory via the three-layer import pipeline (07 §3.1). Use when the user asks to import/migrate documents, a knowledge base, or notes into workspace memory. Layer 1 lands raw documents (zero LLM, instantly searchable); Layer 2 is YOUR distillation work — you read the…