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 QBall-Inc/the-bulwark --skill spec-drift-checkgit clone --depth 1 https://github.com/QBall-Inc/the-bulwarkWrote 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/qball-inc/the-bulwark/spec-drift-check)<a href="https://agentmods.dev/skills/qball-inc/the-bulwark/spec-drift-check"><img src="https://agentmods.dev/badge/skills/qball-inc/the-bulwark/spec-drift-check/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/qball-inc/the-bulwark/spec-drift-check"><img src="https://agentmods.dev/badge/skills/qball-inc/the-bulwark/spec-drift-check.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 | $0.00054 | $0.04234 |
| Opus 5 | $0.00027 | $0.02117 |
| Sonnet 5 | $0.00011 | $0.00847 |
| Haiku 4.5 | $0.00005 | $0.00423 |
Grade A, and why
spec-drift-check 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 11d 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 — 291 lines — stays where its author put it; the contents beside it link to each section on GitHub.
spec-drift-check
Audits a Work Package brief (or any spec document) for drift against the current code state. The skill extracts every verifiable factual claim from the subject spec — file paths, line refs, function and type names, constants, sequence-of-events claims, dependency claims, state claims — verifies each against current code via Grep / Read / Bash, categorizes findings (CONFIRMED, drift variants, AC re-interpretation, GAP) at LOW / MEDIUM / HIGH severity, rewrites the implementation plan based on findings, and emits a structured verdict (PROCEED, PROCEED_ADJUSTED, STOP_USER_APPROVAL). The verified plan supersedes the original spec for the rest of the work package.
When to Use This Skill
Load this skill when the user request matches ANY of these patterns:
| Trigger Pattern | Example User Request |
|---|---|
| Pre-WP verification | "Run spec drift check on the P10.16 brief", "Verify this brief before I start the WP" |
| Drift audit | "Check for drift between the spec and current state", "Audit this brief for stale claims" |
| Claim verification | "Extract claims from this brief and verify them", "Are the file refs in this doc still valid?" |
DO NOT use for:
- Implementing a brief (use the relevant implementer skill / agent — this skill is verification only)
- General code review or PR review (use
code-review) - Test audit (use
test-audit) - Debugging issues (use
issue-debugging)
This skill is READ-ONLY with respect to the subject spec. It does NOT modify the input brief or any code referenced by it. The skill DOES write its own outputs — verification log under $PROJECT_DIR/logs/spec-verify-*.md and diagnostic YAML under $PROJECT_DIR/logs/diagnostics/ — those are not "modifications" of the subject. To FIX issues found in the subject spec, the user invokes a separate skill (manual edits, fix-bug, or an implementer agent). The skill's value is the audit + adjusted plan, not the fix. The frontmatter lists Edit under disallowed-tools, which removes it from the available tool pool while the skill runs, preventing accidental subject-spec modification at the permission layer.
What ships with it
17 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.
- evals/evals.json 1.4 KB
- evals/triggers.json 1.1 KB
- examples/clean-spec.md 2.8 KB
- examples/expected-output-clean.yaml 3.4 KB
- examples/expected-output-high-drift.yaml 3.6 KB
- examples/expected-output-low-drift.yaml 2.3 KB
- examples/high-drift-spec.md 2.6 KB
- examples/low-drift-spec.md 1.8 KB
- references/anti-patterns.md 5.6 KB
- references/output-template.md 6.4 KB
- references/step-1-claim-extraction.md 6.5 KB
- references/step-2-verification-methods.md 7.5 KB
- references/step-3-categorization.md 6.4 KB
- references/step-4-plan-adjustment.md 5.3 KB
- references/step-5-log-template.md 8.1 KB
- references/step-6-decision-matrix.md 6.4 KB
- templates/findings-output.yaml 4.1 KB
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.
- 11d ago First seen · 291 lines · 54 tokens per session scan A 7ccf0b854520
spec-drift-check is a skill published in the GitHub repository QBall-Inc/the-bulwark (8 stars, last pushed today), licensed MIT. It adds 54 tokens to every session and 4,234 once invoked, about $0.0003 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
github-pr-review
Handles PR review comments and feedback resolution. Use when user wants to resolve PR comments, handle review feedback, fix review comments, address PR review, check review status, respond to reviewer, verify PR readiness, review PR comments, analyze review feedback, evaluate PR comments, assess review suggestions, or…
github-pr-creation
Creates GitHub Pull Requests with automated validation and task tracking. Use when user wants to create PR, open pull request, submit for review, or check if ready for PR. Analyzes commits, validates task completion, generates Conventional Commits title and description, suggests labels. NOTE - for merging existing…
adversarial-reviewer
Use when you want a genuinely critical review of recent changes — before merging a PR, after a sprint, or when you suspect the review is being too agreeable. Forces perspective shifts through three hostile reviewer personas that catch blind spots the author's mental model shares with the reviewer. Triggers on…
card-review
Internal skill for onboard's semantic review gate. After the deterministic validation gate (which catches invented symbols / hash drift / secrets), this skill drives a sampled claude -p pass that checks whether the TLDR and role are DIRECTIONALLY ACCURATE given the file's structure and head snippet. The deterministic…
triage-report
Drain the .review/ report queue. Resolves un-triaged scan/audit reports (producer-agnostic — check-docs-consistency and any other scanner), dispatches the triage-report subagent (Sonnet) one report at a time for per-finding dispositions (promote / patch / dup / needs-investigation / dismiss), then the gateway absorbs…
review-ticket
Use as a reviewer agent to review another agent's inreview ticket — never your own. Judge whether each acceptance criterion is genuinely met and the change is sound, then record an ADVISORY verdict (per-AC evidence + an overall RECOMMEND APPROVE / RECOMMEND CHANGES line) via the scoped Dispatch MCP, leaving the ticket…