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-shipgit 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-ship)<a href="https://agentmods.dev/skills/joenandez/spectre/spectre-ship"><img src="https://agentmods.dev/badge/skills/joenandez/spectre/spectre-ship/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-ship"><img src="https://agentmods.dev/badge/skills/joenandez/spectre/spectre-ship.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 | $0.00070 | $0.01675 |
| Opus 5.5 | $0.00028 | $0.00670 |
| Sonnet 5.5 | $0.00014 | $0.00335 |
| Haiku 4.5 | $0.00007 | $0.00168 |
Grade A, and why
spectre-ship 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 1mo 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 — 59 lines — stays where its author put it; the contents beside it link to each section on GitHub.
ship
Own cleanup, rebase, advisory suite, and draft PR.
Inputs
$ARGUMENTS: optional feature root/artifact, target (defaultorigin/main), feedback focus, or--draft; live branch/tree/remotes/PR state.
Feature root
- Reuse a managed
FEATURE_ROOTonly when explicit/current-thread evidence ties it to this work (physical directory wins; never branch/recency/lifecycle/scans); distinct work ignores ambient roots. Otherwise, including on collision, standalone MUST first load and followSkill(spectre-feature-root)through DONE; orchestrated calls escalate. Keep writes beneath it and pass it unchanged.
Proof independence
Proof is optional: do not inspect, infer, invoke, or gate on it. Only authority/publication-safety impasses prevent PR creation.
Outputs + DONE
- Return PR URL, commits, rebase/tested SHA, verification/repairs,
CI: pending, measurement, andPR_OPENEDstatusPASS|REPAIRED|PRE_EXISTING_FAILURES|INDETERMINATE|KNOWN_FAILURES_REMAIN.
DONE when: Prune analysts/Test testers finish their gates, then Sweep, rebase, one full suite, repair/routing, measurement relay/degradation, and PR URL. Non-green verification never prevents DONE.
Method / guardrails
- Resolve once. Confirm branch/target/set; reject
main/master. Read the exact branch; never derive work identity from branch, recency, or ambient roots — selection binds to the frozen post-rebase candidate.FEATURE_ROOTis artifact-only. Reuse unambiguous ancestralCLEANED_THROUGH_SHA. Invokenode "${PLUGIN_ROOT}/hooks/scripts/workflow-cli.mjs" measure start --label Ship; below,measuremeansnode "${PLUGIN_ROOT}/hooks/scripts/workflow-cli.mjs" measure. - Parallel cleanup boundary. Classify P0-P3; invoke
measure startfor Prune/Test. Load${PLUGIN_ROOT}/skills/spectre-ship/references/prune.mdand${PLUGIN_ROOT}/skills/spectre-ship/references/test.mdwith the set/risk/FEATURE_ROOT/--orchestrated. Skill loading imports instructions, not phase delegation: the primary owns both contracts and makes one parallel dispatch of leaf@spectre_analyst/@spectre_testerbatches; no phase child spawns agents. Neither stages/commits. Invokemeasure finishafter all batches, attaching a child identity only when singular; relay compact paths/checks and repair/route cross-boundary needs unlessNEEDS_AUTHORITY. - Sweep. Start
measure start --label "Sweep"; load${PLUGIN_ROOT}/skills/spectre-ship/references/sweep.mdwith--orchestrated, unchanged set, and both results. It alone integrates stale/uncovered checks, repairs attributable failures, and commits; invokemeasure finishfor the Sweep snapshot. - Rebase. Start
measure start --label "Rebase"; load${PLUGIN_ROOT}/skills/spectre-ship/references/rebase.mdwith target,--orchestrated --verification-owner parent; retain backup summary. No checks; invokemeasure finishfor the Rebase snapshot. - Observe one full suite after rebase. At
FULL_SUITE_SHA, derive and freeze candidate{repository, base, head, diff}asEXPECTED_BASE_SHA/EXPECTED_HEAD_SHA/EXPECTED_DIFF_SHA256. Before the first PR side effect, select against that frozen tuple withknowledge-cli.mjs work membership --branch <exact-branch> --candidate <tuple-json>; freeze itsselectedwork-ID set and reportexcluded/ambiguouswithout substituting recency. Still before Create PR, invokeSkill(spectre-work-record)per selected record: finalize a complete one withexecution.statefinalizedandremainingWorkexactlyNone.; refresh a blocked/incomplete one truthfully and never finalize it. No PR, CI, review, or readiness value entersremainingWork. Then invokemeasure start --label "Full suite"andmeasure start --label "Create PR". In parallel run the full-suite lane andSkill(spectre-create_pr)pending with target,--orchestrated --pr-phase pending, the frozen set, and the tuple; it returns a draft with local verificationRUNNING. At each end invokemeasure finishwith child identity when available. No duplicate suites or raw child output.- Attribute failures
branch-caused|unrelated|indeterminate, preferring target-SHA CI and otherwise reproducing only the failing check at target. - Repair branch-caused families; rerun only failing/affected checks, never the full suite. Route unrelated/indeterminate findings; record repaired HEAD and
CI: pending. - Verification is evidence, never a stop condition.
- Attribute failures
- Create PR and measure. After suite call
Skill(spectre-create_pr)with target,--orchestrated --pr-phase final-update, URL/body, the same frozen set, refreshed tuple, andFINAL_VERIFICATION_SUMMARY; change Testing only. Refresh claims after repairs; onPR_CANDIDATE_STALE, refresh and retry. Once PR identity exists, re-runwork membershipagainst the final tuple after any repair; a record it no longer proves is reported, not associated. Annotate the still-selected set through onework associate --work-id <id> … --pull-request-id <id> --candidate <final-json> --pull-request <json>, replayingretry.expectedRevisionsvia--expected-revisionsuntilassociated/noopand reportingpartialremaining. Association annotates only: it never confers finality, PR state isdraft-open, never merged, and a later Execute run on this branch keeps its own record and joins this same PR only when the updated candidate proves membership. Ambiguity or capture failure returns recovery input but does not block PR. Runmeasure summary --rows … --outer-snapshot … --persist --project-dir … --feature-root "$FEATURE_ROOT" --base-sha "$EXPECTED_BASE_SHA" --head-sha "$EXPECTED_HEAD_SHA" --diff-sha256 "$EXPECTED_DIFF_SHA256"; relay table, persistence status, and history path. Persistence degradation never blocks PR completion; never inspect transcripts, track clocks, or calculate. Unattributed Prune/Test and Full suite/Create PR use one exact parallel-group total; unavailable measurement never blocks Ship.
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.
- 1mo ago Changed 3719023e9f83
- 1mo ago Changed 53f0d0966311
- 1mo ago Changed · -1 lines 44b6d77d8132
- 1mo ago Changed · +6 lines 748a9363761e
- 2mo ago Changed · +1 lines · -1 tokens per session e5710eb627cd
- 2mo ago First seen · 53 lines · 71 tokens per session scan A c9d31ffda9bc
spectre-ship is a skill published in the GitHub repository joenandez/spectre (164 stars, last pushed 1mo ago), licensed MIT. It adds 70 tokens to every session and 1,675 once invoked, about $0.0003 per session on Opus 5.5. 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.
spectre-ship compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| spectre-shipjoenandez/spectre | Skill | 164 | 1mo ago | MIT | A | 70 | Claude Code, Codex |
| merge-batchdiegosouzapw/OmniRoute | Skill | 83k | 6d ago | MIT | A | 46 | Claude Code |
| prowler-commitprowler-cloud/prowler | Skill | 15k | yesterday | Apache-2.0 | A | 44 | Claude Code, Codex |
| gh-auth-isolationgithub/gh-aw | Skill | 5.6k | yesterday | MIT | A | 20 | Claude Code, Codex |
| github-skillzeenie-ai/OpenCompany | Skill | 1.1k | yesterday | MIT | A | 51 | Claude Code |
| comet-githubrpamis/comet | Skill | 3.3k | yesterday | MIT | A | 72 | Claude Code, Codex |
| release-helperSahir619/fable-method | Skill | 2.3k | 1mo ago | MIT | A | 43 | Claude Code |