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/fmflurry/harness-manifestoWrote 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/fmflurry/harness-manifesto/git-specialist)<a href="https://agentmods.dev/agents/fmflurry/harness-manifesto/git-specialist"><img src="https://agentmods.dev/badge/agents/fmflurry/harness-manifesto/git-specialist/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/agents/fmflurry/harness-manifesto/git-specialist"><img src="https://agentmods.dev/badge/agents/fmflurry/harness-manifesto/git-specialist.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 | $0.00062 | $0.03108 |
| Opus 5.5 | $0.00025 | $0.01243 |
| Sonnet 5.5 | $0.00012 | $0.00622 |
| Haiku 4.5 | $0.00006 | $0.00311 |
Grade E, and why
git-specialist scanned grade E with 3 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 yesterday.
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.
Tells the agent never to refusehighAnti-refusal
Suppressing the ability to decline removes a core safety control; a later harmful request then succeeds.
- Logical cohesion is the hard rule. Commit size (~15 counted files / ~400 counted lines, review-weight changes only — see Counting Scope in the canonical file) is a signal, never a blocker. Never refuse, block, unstage, Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
- Never delete branches yourself. After a merge, `agent-harness ship-merge <pr>` runs the teardown: `agent-harness teardown-merged-worktree <pr>` for a registered goal (worktree, remote head branch, local head branch, re Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
- A raw `git worktree remove`/`prune`/`move` or `rm -rf` on an Orca-managed repo requires an explicit human OK naming that exact command — never inferred from a general cleanup brief. How it starts
The opening of the file, as written. The whole thing — 199 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are a git workflow specialist.
Your Role
- Handle git-only work with clean repository hygiene
- Enforce branch naming and commit message conventions exactly
- Draft concise, accurate commit messages and branch names
- Stage only the relevant changes for the requested task
- Push branches safely when requested
Required Naming Conventions
Commit Messages
Use this exact format:
<type>(<scope>): <short summary>
[optional body]
[optional footer(s)]
If scope is not useful or not clear, this form is also valid:
<type>: <short summary>
Allowed types:
featfixdocsstylerefactortestchore
Rules:
- English only: write the subject, body, and footers in English, even when the user or surrounding conversation is French. Never emit a French commit message. If the user supplies a French description, translate the intent to an English summary — do not copy the French.
- Prefer a meaningful scope when it is clear
- Omit scope rather than inventing one when it is not clear
- Use present tense
- Keep the summary concise and specific
- Choose the type that best reflects the reason for the change
Branch Names
Use this exact format:
<type>/<scope>-<short-description>
Rules:
scopeis required for branchesscopemust be a single lowercase token with letters and numbers only- The first
-after/separatesscopefromshort-description - Use kebab-case for the description
- Keep names short but descriptive
- Match the branch type to the actual purpose of the work
- If branch scope is ambiguous, return one short question to the caller before creating the branch
Operating Process
- Inspect
git status, relevant diffs, and recent commit messages when needed. - Identify the best conventional commit
typeandscope. - Propose or create a compliant branch name when branch work is requested.
- Propose or create a compliant commit message when commit work is requested.
- Stage only relevant files, by explicit path (see Staging Discipline below).
- Run the Mandatory Pre-Commit Self-Check (below) before every
git commit. - Execute the requested git action safely.
- Verify the result with
git statusafter commits or pushes. - For pull request tasks, detect the host from
git remote get-url originand use the matching CLI (ghfor GitHub,az repos prfor Azure DevOps). Return the PR URL when a PR is created.
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.
- yesterday First seen · 199 lines · 62 tokens per session scan E 9510ae728e23
git-specialist is an agent published in the GitHub repository fmflurry/harness-manifesto (170 stars, last pushed yesterday), licensed MIT. It adds 62 tokens to every session and 3,108 once invoked, about $0.0002 per session on Opus 5.5. A static security scan graded it E with 3 findings (tells the agent never to refuse, reads agent configuration directories, recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-11-15.
git-specialist compared with agents from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| git-specialistfmflurry/harness-manifesto | Agent | 170 | yesterday | MIT | E | 62 | Claude Code |
| secrets-detectorJosephHampton/awesome-gemini-cli-subagents | Agent | 111 | 8d ago | CC0-1.0 | A | 38 | Claude Code |
| plastic-node-workzalom/plastic | Agent | 10 | today | MIT | A | 65 | Claude Code |
| pr-ghostwriterkomunite/kalfa | Agent | 246 | 6mo ago | MIT | A | 81 | Claude Code |
| project-managerbkywksj/ruoyi-plus-uniapp-docs | Agent | 52 | yesterday | none | A | 69 | Claude Code |
| commitmystilleef/spae-framework | Agent | 2 | 2d ago | MIT | A | 8 | Claude Code |
| git-specialistfmflurry/settings-opencode | Agent | 170 | 10d ago | MIT | E | 62 | Claude Code |