Borrowing it
Nothing to install: this file belongs to open-software-network/os-clovy. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/open-software-network/os-clovy/main/.agents/skills/repo-delegate/SKILL.mdgit clone --depth 1 https://github.com/open-software-network/os-clovyWrote 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/open-software-network/os-clovy/repo-delegate)<a href="https://agentmods.dev/skills/open-software-network/os-clovy/repo-delegate"><img src="https://agentmods.dev/badge/skills/open-software-network/os-clovy/repo-delegate.svg" alt="Measured on agentmods" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to high
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- high Agent Snooping · line 70 Skill reads from agent configuration directories (.claude/, .codex/, .gemini/). These directories may contain API keys, personal settings, and other credentials that the skill has no legitimate need to access.Fix: Remove all code or instructions that access agent configuration directories (.claude/, .codex/, .gemini/). If configuration values are needed, pass them explicitly as parameters or environment variabl
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 | $0.00112 | $0.01944 |
| Opus 5 | $0.00056 | $0.00972 |
| Sonnet 5 | $0.00022 | $0.00389 |
| Haiku 4.5 | $0.00011 | $0.00194 |
Grade B, and why
repo-delegate scanned grade B with 1 finding 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.
Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
**Codex model/effort/speed** (defaults come from `~/.codex/config.toml`; How it starts
The opening of the file, as written. The whole thing — 141 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Repo delegate
Hand a well-scoped implementation task to another harness and get a
report back. Same structure as repo-review: one prompt template
(PROMPT.md), one runner script per harness, a template for
adding more. The delegate edits files and runs checks; the caller reviews
the diff and commits — delegates never touch git state.
When
- A fix round produced precise per-finding prescriptions (repo-review loop).
- Bulk implementation against a written contract (repo-build-pr delegation).
- A second, independent implementation attempt from a different model family.
Not for: architecture, contracts, go/no-go calls — those stay with the orchestrating model (see repo-build-pr's model orchestration rules).
How
-
Write the task brief to a file. Like a contract: exact scope and file ownership, the interface to build against, repo conventions to follow, and an instruction to report deviations instead of improvising. If the brief would be longer than the diff, don't delegate. For a defect fix, make the gate fail first: encode the bug as a failing regression check (test or fixture assertion) before dispatch and say so in the brief — the delegate's "gate passed" is then proof the defect died, not a claim.
State invariants, not failure-path branches. A brief that dictates what to do when something goes wrong ships the orchestrator's unexamined premise straight into the code, and the delegate will implement it faithfully. Say "never type into an unintended app"; do not say "on timeout post the keystroke anyway so text is not lost". Before writing any "or else X is lost" rationale, verify X can actually be lost. In PR #676 it could not — the transcript was already on the clipboard and in history — so the brief's own safety clause recreated the exact bug the PR existed to fix, and only a cross-harness adversarial run caught it.
Mutation-check any fixture the delegate authored. A test written by the same agent that wrote the fix can pass for reasons unrelated to the fix. Revert the fix, confirm the test goes red, restore. PR #676: a delegate's "notice does not fire" test asserted the absence of a label in a state where an unrelated guard already prevented it — green with the fix removed. PR #677 is the harder version: a delegate rewrote an existing test so its own regression would pass.
-
Fill and dispatch (
-tis the brief file; gate defaults topnpm check && pnpm typecheck && pnpm test):
What ships with it
5 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 · 141 lines · 112 tokens per session scan B 7240d49a1163
repo-delegate is a skill published in the GitHub repository open-software-network/os-clovy (357 stars, last pushed 7d ago), licensed MIT. It adds 112 tokens to every session and 1,944 once invoked, about $0.0006 per session on Opus 5. A static security scan graded it B with 1 finding (reads agent configuration directories). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
personal-commands
Edit hex.config.ts, not .hex-sdk. Config and installed dependencies execute with the user's normal permissions; use only trusted packages and avoid network, filesystem, environment, or subprocess access unless the requested command requires it. Run bun run check after editing.
hex-personal-commands
Configure HEX personal voice commands, voice-dictation control phrases, and deterministic transcript transformations. Use when the user wants to add or change commands or transformations for the HEX macOS voice app.
macparakeet-stt
Use when the user asks to transcribe audio or video, search MacParakeet's local speech library, retrieve cited transcript segments or knowledge cards, or inspect and manage saved meeting artifacts on macOS Apple Silicon.
murmur-mcp
Query Murmur's local read-only MCP server for gated, indexed knowledge — semantic + full-text search over meetings and notes, entity dossiers, structured transcripts, open-commitment rollups, and typed note-folder database queries. Prefer this over grepping raw vault files. Use when you need to find, retrieve, or…
murmur-vault
How to coexist with Murmur's notes in an Obsidian vault — note anatomy, front-matter keys, managed sections you must never edit, wikilink discipline, entity stubs, where files live, the locked-folder disappeared-file rule, and the add-freely/prefer-append coexistence contract. Use when reading, editing, or creating…
frame-macos-notification
Realistic macOS notification banner with app icon, title, and body, suited to video overlays or product teasers.