Borrowing it
Nothing to install: this file belongs to MarkusNeusinger/cite-citadel. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/MarkusNeusinger/cite-citadel/main/.claude/skills/verify-corpus/SKILL.mdgit clone --depth 1 https://github.com/MarkusNeusinger/cite-citadelWrote 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/markusneusinger/cite-citadel/verify-corpus)<a href="https://agentmods.dev/skills/markusneusinger/cite-citadel/verify-corpus"><img src="https://agentmods.dev/badge/skills/markusneusinger/cite-citadel/verify-corpus.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 | $0.00539 | $0.10938 |
| Opus 5 | $0.00269 | $0.05469 |
| Sonnet 5 | $0.00108 | $0.02188 |
| Haiku 4.5 | $0.00054 | $0.01094 |
Grade C, and why
verify-corpus scanned grade C with 1 finding 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 6d 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.
Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
survives the `rm -rf`. How it starts
The opening of the file, as written. The whole thing — 605 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Verify a corpus end-to-end
Ten shipped corpora, each a corpora/<name>/ bundle (raw/, sometimes stages/ or a
materializable repo tree, a README.md) plus a hidden answer key at
.claude/skills/verify-corpus/<name>/ground-truth.md. The ingest agent
never sees the key — it lives outside the corpus, and Mode A points CITADEL_RAW_DIR at the
corpus raw/ only (defense in depth). This skill runs the real pipeline into a sandbox, then
grades that sandbox wiki against the key. Grading is a two-phase FACTS-style gate: phase 1 =
citadel check + lint exit 0 (structural eligibility); phase 2 = answer-key content grading.
Modes A/B grade the ingest lifecycle. Mode C (below) grades the curate lifecycle end-to-end — it seeds known defects into a sandbox, proves the offline detectors flag exactly them, then runs real curate sessions and grades that the model FIXES them without breaking valid pages.
Usage: verify-corpus <beverages|kelvarra|leuchtfeuer|pemberley|injection-resistance|clockwork|flurfunk|gazette|kontor|werkhof|all> [--grade-only]
| corpus | what it stresses | sandbox note | ground-truth |
|---|---|---|---|
beverages |
organized / links / provenance on a messy coffee+tea corpus; the showcase wiki | 10 files, one pass each | .claude/skills/verify-corpus/beverages/ground-truth.md |
kelvarra |
the hardest guarantee: a fictional world stated wrong about reality, kept as-stated and cited, never corrected | 7 files, one pass each | .claude/skills/verify-corpus/kelvarra/ground-truth.md |
leuchtfeuer |
reconcile / delete / force across 3 dated waves; temporal supersession; German→English; opinions & style | 3 waves (see the wave protocol) | .claude/skills/verify-corpus/leuchtfeuer/ground-truth.md |
pemberley |
large-source multi-segment chunking; relationship extraction; in-novel misinformation; narrative supersession | one ~730k-char file → ~18 segments, HOURS — set a LONG timeout and force chunking (see the pemberley note) | .claude/skills/verify-corpus/pemberley/ground-truth.md |
injection-resistance |
embedded adversarial instructions treated as content, never executed; real facts still extracted | 3 files, 3 quick sessions | .claude/skills/verify-corpus/injection-resistance/ground-truth.md |
clockwork |
a whole git repo folded in as ONE digest (kind=repo), then a second commit driving kind=repo-reconcile; folder-keyed provenance; one documented default superseded |
materialize the repo + 2 commits (see the clockwork note) | .claude/skills/verify-corpus/clockwork/ground-truth.md |
flurfunk |
informal genres (chat / social / interview / application / forum); attribution ("X said Y" ≠ "Y is true"), in-thread reversal, a quote-tweet negative row, CV timeline | 7 files, one pass each | .claude/skills/verify-corpus/flurfunk/ground-truth.md |
gazette |
PDF sources: CITADEL_PDF_MODE text-vs-images (a figure-only number + an image-only page), the publications genre, references-are-not-sources, page locators, and chunked-window edges |
6 files (5 PDFs + 1 md); three runs (text, images, then a chunked run — see the gazette note) | .claude/skills/verify-corpus/gazette/ground-truth.md |
kontor |
binary Office documents (OOXML .pptx/.docx/.xlsx + legacy OLE .doc/.ppt/.xls); the Office extraction path, an embedded-chart image delta (CITADEL_IMAGE_SUPPORT), dedup-by-basename, ignore-patterns; the usual judgment traps | 11 files (8 Office + 3 junk); two runs (images off then on — see the kontor note) | .claude/skills/verify-corpus/kontor/ground-truth.md |
werkhof |
uniform enumerable sources (a 28-machine register, a 20-code fault catalogue, a 15-customer CSV): Registry completeness (no entry compressed away), per-row citations + locators, promotion past the granularity floor, a status supersession, near-miss keys |
4 files, one pass each | .claude/skills/verify-corpus/werkhof/ground-truth.md |
What ships with it
10 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.
- beverages/ground-truth.md 26 KB
- clockwork/ground-truth.md 15 KB
- flurfunk/ground-truth.md 16 KB
- gazette/ground-truth.md 18 KB
- injection-resistance/ground-truth.md 20 KB
- kelvarra/ground-truth.md 21 KB
- kontor/ground-truth.md 12 KB
- leuchtfeuer/ground-truth.md 33 KB
- pemberley/ground-truth.md 26 KB
- werkhof/ground-truth.md 6.6 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.
- 6d ago First seen · 605 lines · 539 tokens per session scan C 885339d3b88d
verify-corpus is a skill published in the GitHub repository MarkusNeusinger/cite-citadel (2 stars, last pushed 5d ago), licensed MIT. It adds 539 tokens to every session and 10,938 once invoked, about $0.0027 per session on Opus 5. A static security scan graded it C with 1 finding (recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
vue-component-testing
Applies the three-tier test taxonomy for Vue 3 applications: writes unit tests for composables and Pinia stores with Vitest, component tests for behaviour and user interactions with @testing-library/vue, and acceptance tests for full user flows with Playwright. Ensures tests focus on observable behaviour, not…
tmux-real-user-testing
This skill should be used when the user asks to "用 tmux 做真实测试", "保存 tmux 日志", "像真实用户一样测试 Qwen", "生成可复查的 TUI 测试报告", "测试 slash command 交互", or requests a tmux-based real user E2E run with complete readable logs. It guides real TUI usage with step-by-step capture-pane snapshots rather than ANSI raw pipe logs.
e2e-testing
Guide for running end-to-end tests of the Qwen Code CLI, including headless mode, MCP server testing, and API traffic inspection. Use this skill whenever you need to verify CLI behavior with real model calls, reproduce user-reported bugs end-to-end, test MCP tool integrations, or inspect raw API request/response…
terminal-capture
Automates terminal UI screenshot testing for CLI commands. Applies when reviewing PRs that affect CLI output, testing slash commands (/about, /context, /auth, /export), generating visual documentation, or when 'terminal screenshot', 'CLI test', 'visual test', or 'terminal-capture' is mentioned.
agent-reproduce-align
Use after a Codex or Claude Code feature has been implemented in Qwen Code to run the selected reference agent and Qwen Code under the same scenario, capture HTTP and terminal traces, compare request bodies, tool/function schemas, outputs, and iterate until the reproduced behavior is close enough.
test-app
Verify a running Mendix app in a browser with Playwright, with OQL for data assertions. Use when asked to test or validate the app end to end, or to confirm that generated pages actually render.