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 joenandez/spectre --skill spectre-cleangit clone --depth 1 https://github.com/joenandez/spectreWrote 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/joenandez/spectre/spectre-clean)<a href="https://agentmods.dev/skills/joenandez/spectre/spectre-clean"><img src="https://agentmods.dev/badge/skills/joenandez/spectre/spectre-clean/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/joenandez/spectre/spectre-clean"><img src="https://agentmods.dev/badge/skills/joenandez/spectre/spectre-clean.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00085 | $0.01180 |
| Opus 5 | $0.00043 | $0.00590 |
| Sonnet 5 | $0.00017 | $0.00236 |
| Haiku 4.5 | $0.00009 | $0.00118 |
Grade A, and why
spectre-clean 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 today.
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 — 66 lines — stays where its author put it; the contents beside it link to each section on GitHub.
clean
End-to-end cleanup utility. The primary owns scope, risk assessment, sequencing, and synthesis. It is not a Ship dependency; phase work uses spectre-prune, spectre-test, and spectre-sweep.
Inputs
$ARGUMENTS— optional scope hint: commit/SHA,unstaged/staged,context/session, task dir, files, or--orchestratedwhen a parent workflow owns the next step.FEATURE_ROOT— explicit feature name/root or one descendant feature artifact.
Working Set
- Reuse a managed
FEATURE_ROOTonly when explicit/current-thread evidence ties it to this work (physical directory wins; never branch/recency/lifecycle/scans); distinct work ignores ambient roots. Otherwise, including on collision, standalone MUST first load and followSkill(spectre-feature-root)through DONE; orchestrated calls escalate. Keep writes beneath it and pass it unchanged. - Resolve the same full working set used by prune/test/sweep from committed changes, staged, unstaged, and untracked files. If a provided ref/scope is invalid or ambiguous, stop and ask.
- Keep the resolved file set and primary's compact P0-P3 risk plan in-thread; write no working-set/lifecycle artifact.
Outputs + DONE
- Parallel prune and one test-lead result; manual-review/cross-boundary items preserved.
- Primary P0-P3 risk plan, compact check results, one Sweep verification/commit boundary, and
CLEANED_THROUGH_SHA. - Final report: prune/manual review · risk tiers · tests/affected checks · routed findings · sweep commits ·
CLEANED_THROUGH_SHA·NEEDS_AUTHORITY, if any. - DONE when: every phase ran in its owning subagent, the primary supplied risk tiers, repairable findings were repaired/routed by their owner, manual-review items surfaced, and sweep committed or returned genuine
NEEDS_AUTHORITY.
Method / guardrails
- Resolve scope once. Establish files and feature root. Keep dynamic details out of the prompt body; read them live.
- Primary risk assessment. Classify every resolved file P0-P3:
- P0: auth/payment/security/crypto/session/token, PII/permissions/user-data mutation, external handlers, DB migrations,
@critical. - P1: feature components, API/state/business logic, fetch/cache, user-visible errors.
- P2: exported real-logic utilities, validators, transformers, adapters, hooks.
- P3: docs/styles/config/types/constants/barrels/pass-throughs/generated files.
Keep a compact plan in-thread:
- [P{tier}] {file}: {behavior or SKIP reason}.
- P0: auth/payment/security/crypto/session/token, PII/permissions/user-data mutation, external handlers, DB migrations,
- Parallel phases. In one dispatch, start one prune lead for
Skill(spectre-prune)and one test lead forSkill(spectre-test)with the unchanged resolved set,{FEATURE_ROOT}, and the risk plan. Each uses--orchestrated; the test lead owns any internal tester batching. They return compact changed-path/check results, manual-review items, and cross-boundary needs; neither stages nor commits. - Sweep phase. Dispatch a sweep lead with the unchanged set and compact phase results.
Skill(spectre-sweep)runs only stale/uncovered integrated checks, repairs attributable failures, and is the sole pre-rebase commit owner. - Synthesize. Route repairable/cross-boundary findings to their owner; report final state, the Sweep commit
CLEANED_THROUGH_SHA, routed findings, and genuine authority/safety impasses.
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.
- today Changed · +5 lines 158413682d08
- 5d ago Changed · -3 lines · -4 tokens per session c9b409b5c8ef
- 11d ago First seen · 64 lines · 89 tokens per session scan A e799792203e1
spectre-clean is a skill published in the GitHub repository joenandez/spectre (161 stars, last pushed yesterday), licensed MIT. It adds 85 tokens to every session and 1,180 once invoked, about $0.0004 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 2026-08-30.
Other skills, from other repositories
agent-merge-conflict-arbiter
Neutral arbiter for merge conflicts between two agents.
git-advanced-workflows
Master advanced Git workflows including rebasing, cherry-picking, bisect, worktrees, and reflog to maintain clean history and recover from any situation. Use when managing complex Git histories, collaborating on feature branches, or troubleshooting repository issues.
bisect
Bisect regressions or broken builds in the Chromium repository using deterministic binary search, checking out midpoint commits, syncing dependencies with gclient sync, compiling the target with autoninja, and querying the user for test results until the culprit commit is isolated.
mantis-critic
Assesses the production viability of findings, filtering out debug-only features and assertion traps. Use when findings have been validated and you need to confirm they are triggerable in production release builds (with assertions disabled). Don't use for writing reproduction scripts or patches.
hotpath_bump
Bump the hotpath version number across the workspace and related files. Updates crate versions in Cargo.toml files (exact patch version) and version references in the backend middleware, hotpathinit skill, and README (major.minor only). Use when the user wants to bump, bump the version, or release a new hotpath…
comet-github-issue-fix
A workflow for fixing a confirmed Comet-related GitHub issue or review blocker within an isolated scope. It covers matching the work to the relevant workflow, testing the change, checking runtime results, and preparing a careful handoff.