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 AtlasOmnia/hermes-custom-pack --skill coding-worktree-recoverygit clone --depth 1 https://github.com/AtlasOmnia/hermes-custom-packWrote 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/atlasomnia/hermes-custom-pack/coding-worktree-recovery)<a href="https://agentmods.dev/skills/atlasomnia/hermes-custom-pack/coding-worktree-recovery"><img src="https://agentmods.dev/badge/skills/atlasomnia/hermes-custom-pack/coding-worktree-recovery.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 | $0.00056 | $0.05733 |
| Opus 5 | $0.00028 | $0.02867 |
| Sonnet 5 | $0.00011 | $0.01147 |
| Haiku 4.5 | $0.00006 | $0.00573 |
Grade A, and why
coding-worktree-recovery 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.
This is a copy
100% identical to coding-worktree-recovery — 0 lines differ, which has more behind it and is treated as the original. This page carries a canonical link to it rather than competing with it.
How it starts
The opening of the file, as written. The whole thing — 278 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Coding Worktree Recovery
Use this skill when coding-agent work is interrupted, an agent exits without a clean commit, multiple controllers target the same checkout, or the checkout produces inconsistent file/Git behavior.
This skill is about preserving valid work and restoring one trustworthy writer. It does not authorize pushing, merging, publishing, or destructive cleanup.
Core invariants
- One checkout, one writer. Read-only reviewers may overlap; write-capable agents may not.
- The live filesystem is authoritative. Agent summaries and successful process exits are evidence leads, not proof.
- Never clean another writer's work. Do not reset, stash, checkout, delete, or overwrite unexplained changes.
- Commits are checkpoints, not completion. Re-run controller gates after the final commit.
- Mechanical status and reasoning are separate. A deterministic watchdog may report process/Git state; acceptance still requires a controller review.
Recovery workflow
1. Freeze new writers
- Do not launch another implementation agent yet.
- Enumerate all live coding/Hermes processes whose full command references the checkout.
- Record PID, parent PID, process group, start time, elapsed time, and command.
- Inspect child processes as well as tool-managed process handles; an interrupted parent can leave an orphan writer.
- Trace workers back to long-lived TUI/gateway controllers. Killing only a visible
hermes ... chatPID is insufficient when its controller can respawn or resume another worker. - Match writers by exact command semantics (for example,
hermes ... chatplus checkout), not by broad words such ashermes, model names, or the checkout path; broad matching misclassifiestsserverand other read-only services. - If a stale TUI or messaging controller repeatedly respawns workers, preserve the user session while stopping its already-spawned worker/monitor process groups. Trace the exact originating conversation before transferring ownership.
- Do not assume
/stopcancels prior task-owned background work. It may correctly report no active task while a previously launched worker survives or its completion notification triggers a successor. When that happens, send an explicit plain-language cancellation instruction into the exact originating conversation, require acknowledgement, and independently verify that no new generation appears. Do not suspend the production messaging gateway for this purpose.
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 · 278 lines · 56 tokens per session scan A 6dbc12c54627
coding-worktree-recovery is a skill published in the GitHub repository AtlasOmnia/hermes-custom-pack (56 stars, last pushed 23d ago), licensed MIT. It adds 56 tokens to every session and 5,733 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. It is 100% identical to coding-worktree-recovery, differing in 0 lines, and is treated as a copy.
Other skills, from other repositories
dependency-update-bot
Scans your project for outdated npm, pip, Cargo, Go, or Ruby packages. Runs a CVE security audit. Fetches changelogs, summarizes breaking changes with Gemini, and opens one PR per risk group (patch, minor, major). Includes Diagnosis Mode for install conflicts. Use when asked to update dependencies, check for outdated…
explain-this-pr
Takes a GitHub PR URL or the current branch and writes a plain-English explanation of what it does and why, then posts it as a PR comment. Use when asked to explain a PR, summarize a pull request, write a plain-English description of a PR, add a summary comment to a PR, or understand what a PR changes. Trigger when a…
pr-description-writer
Use when the user asks to write, draft, generate, or update a GitHub pull request description for the current branch. Reads the branch diff and commit messages, writes a structured PR body (summary, changes, testing), and optionally creates or updates the PR via the gh CLI.
github-pre-push-gates
Before pushing a branch to a shared remote, run through these gates. They prevent pushing credentials, PII, private history, or a broken tree. The skill covers the pre-PR quality phase — use github-pr-workflow for the PR lifecycle itself.
coding-worktree-recovery
Use this skill when coding-agent work is interrupted, an agent exits without a clean commit, multiple controllers target the same checkout, or the checkout produces inconsistent file/Git behavior.
local-app-github-publishing
Use this when the user asks to push a local app/prototype/project to GitHub for the first time, especially Electron, iOS/Xcode, mobile, desktop, or AI/API-integrated apps.