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 agentmods add skills/kimmingul/xllm/scribenpx skills add kimmingul/xllm --skill scribegit clone --depth 1 https://github.com/kimmingul/xllmWrote 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/kimmingul/xllm/scribe)<a href="https://agentmods.dev/skills/kimmingul/xllm/scribe"><img src="https://agentmods.dev/badge/skills/kimmingul/xllm/scribe.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 | $0.00090 | $0.00731 |
| Opus 5 | $0.00045 | $0.00365 |
| Sonnet 5 | $0.00018 | $0.00146 |
| Haiku 4.5 | $0.00009 | $0.00073 |
Grade A, and why
scribe scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 65 lines — stays where its author put it; the contents beside it link to each section on GitHub.
scribe — Cheap-prose lane for git chores (xllm)
Split of labor: deterministic collectors gather the diff/log (no LLM), the cheapest healthy advisor writes only the prose, and you execute the git command. push/tag are pure mechanics — no model call at all.
Resolve the script
Claude Code: "${CLAUDE_PLUGIN_ROOT}/scripts/xllm-scribe.js".
Other hosts: <plugin-root>/scripts/xllm-scribe.js (two directories above
this SKILL.md).
Run (Bash tool)
node <xllm-scribe.js> commit # staged diff → commit message
node <xllm-scribe.js> pr --base main # commits+diffstat → PR title/body
node <xllm-scribe.js> release --from v0.5.0 # log range → release notes
node <xllm-scribe.js> notes --from v0.5.0 # log range → changelog entry
# all accept --provider <spec> to override routing
Message arrives on stdout (diagnostics on stderr), so:
MSG=$(node <xllm-scribe.js> commit) && git commit -m "$MSG"
Contract
- Routing:
scriberole — local-first (free),release/notesauto-escalate to a cloud model when one is healthy (narrative needs judgment). Pin per project:--set-role scribe ollama:qwen3.6:latest@low. - Output is validated deterministically (Conventional Commits format, subject ≤72 chars, no fences, no invented facts instruction) with one corrective retry; exit 3 means validation failed — show the raw text to the user for review instead of using it blindly.
- Privacy: the diff goes only to the routed advisor and is never persisted by xllm (no artifact) — the git object itself is the record. For sensitive repos keep scribe on local models.
- Show the generated message to the user (or apply directly if they asked you to commit), then run the git command yourself. xllm never runs git.
- Advisors stay read-only throughout — this lane generates text, nothing else.
When NOT to use
- push / tag creation → pure git commands; no model needed at all.
- Release strategy, breaking-change judgment, "should we ship" → that is
analysis, not scribing; use
ask/review roleswith a strong model. - If no advisor is healthy: on Claude Code, spawn a host-native cheap-model
subagent instead (Agent tool with
model: haiku— pass it the same collector output and template contract); otherwise write the prose yourself. Never block the user on a missing advisor.
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.
- 3d ago First seen · 65 lines · 90 tokens per session scan A 719f7b85cd6b
scribe is a skill published in the GitHub repository kimmingul/xllm (2 stars, last pushed 11d ago), licensed MIT. It adds 90 tokens to every session and 731 once invoked, about $0.0005 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 2026-08-31.
Other skills, from other repositories
changelog
Record shipped work in the monthly changelog (append-only). Use when something ships or when asked "what changed", "add to changelog", "release notes", or "what shipped this month".
handoff
Ship completed work: verify locally, commit related changes, push, create or update the PR, watch CI/reviews, and fix until merge-ready or escalated. Use for "ship it", "create PR", "handoff", or finished code needing delivery.
arn-code-ship
This skill should be used when the user says "ship it", "arness code ship", "create PR", "open pull request", "push and PR", "commit and push", "wrap up", "ship the feature", "ship the fix", "ready to ship", "push changes", "finalize", "finish up", or wants to commit, push, and optionally open a pull request. Guides…
arn-shipping
This skill should be used when the user says "shipping", "arness shipping", "ship it", "ship", "create PR", "open pull request", "push and PR", "commit and push", "wrap up", "ship the feature", "ship the fix", "ready to ship", "push changes", "finalize", "finish up", "done", "arn-shipping", or wants to commit, push…
smart-git-commit
Use this skill for ANY git operation — commits, pushes, PRs, releases, or version tagging. Triggers on: "commit", "push", "save my changes", "create a PR", "open a pull request", "ship this", "make a release", "tag this version", "checkpoint my work", or any request to record or publish code changes. Produces…
git-flow
Use when creating a branch, using worktrees for parallel work, committing, or wrapping up. Covers branch naming, conventional commits, short-lived branches, worktree isolation, and final merge/cleanup.