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/knowhowlab/agent-testkit/test-manualnpx skills add knowhowlab/agent-testkit --skill test-manualgit clone --depth 1 https://github.com/knowhowlab/agent-testkitWrote 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/knowhowlab/agent-testkit/test-manual)<a href="https://agentmods.dev/skills/knowhowlab/agent-testkit/test-manual"><img src="https://agentmods.dev/badge/skills/knowhowlab/agent-testkit/test-manual.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 | $0.00089 | $0.01464 |
| Opus 5 | $0.00044 | $0.00732 |
| Sonnet 5 | $0.00018 | $0.00293 |
| Haiku 4.5 | $0.00009 | $0.00146 |
Grade A, and why
test-manual scanned grade A 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 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.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
not execute build, run, curl, kill, docker, or any stack-exercising How it starts
The opening of the file, as written. The whole thing — 127 lines — stays where its author put it; the contents beside it link to each section on GitHub.
test-manual — Conduct the manual plan
Goal: Walk the user through TESTS_MANUAL.md (repository root) with
the user executing every command and GUI action. Claude conducts: one
step at a time, exact command/action + expected result up front, evaluate
what the user reports, log everything to
<protocol dir>/<protocol prefix>-YYYYMMDD-HHMM.md (UTC timestamp at
start; create the directory if missing).
Pre-flight
- Locate
TESTS_MANUAL.mdat the repo root (or nearest ancestor). Missing → say so and offer to runtest-init. Do not improvise. - Read the plan: Meta, Scopes, Facts, Fixtures, Phases, Wrap-up.
- Resolve arguments against
## Scopes; unknown scope → ask. The user may also name phases directly ("just Phase 4 and 5").
Critical rules (inverted from test-e2e)
- The user runs the commands and drives the GUI — never Claude. Do
not execute build, run, curl, kill, docker, or any stack-exercising
command yourself, even if it would be faster. Present each shell
command copy-paste-ready and suggest the
!prefix (! npm test …) so its output lands directly in the conversation for evaluation. Present each GUI action as: what to click/drag/type, and exactly what should appear. - One step at a time. Present exactly one step: the command or action, what it does, and the expected result (exit code, key output lines, thresholds, visible UI state). Then stop and wait. Never dump a phase's worth of steps at once.
- Evaluate, don't assume. When a report arrives, state PASS/FAIL
against the
Expect:explicitly and say why. Ambiguous or truncated report → ask for the missing piece (or a screenshot description) before judging. Check## Factsbefore judging — documented caveats are not failures. - Propose fixes, apply only on instruction. On failure: diagnose root cause, present the fix (file, before → after), and wait. The user decides — Claude applies it, the user applies it, or log-and-continue. Never edit code unasked.
- The user may skip. Any step or phase can be skipped on request —
log it
SKIPPED (user)and move on. Missing prerequisites →SKIPPED (prereq), recording which. - Claude keeps the books. Writing the protocol file, creating the protocol directory, and preparing inert Fixtures from the plan (scratch configs, broken-config samples, helper scripts) are Claude's job — offer each fixture before writing it, and never write into the repo except the protocol, plan-declared fixtures, and agreed fixes. Everything that executes is the user's.
- Retry guidance, not retry execution. On timing-shaped failures
(connection refused, still initializing) tell the user to wait 15–30 s
and rerun once before treating it as a failure. Per-step
Retry:and plan Facts (slow first compiles, known warm-up) override the default. - Destructive steps are announced. A step marked
Destructive:is presented with its consequence stated plainly ("this wipes the app's cached state — binary selection will be gone") before the user runs it.
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 · 127 lines · 89 tokens per session scan A 68ab55682805
test-manual is a skill published in the GitHub repository knowhowlab/agent-testkit (3 stars, last pushed 1mo ago), licensed MIT. It adds 89 tokens to every session and 1,464 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
watch
File sentinel that monitors the working directory for changes and marker comments, then auto-triggers appropriate skills. Poll-based via git diff against the last scan commit. Writes intake items for batch processing and routes marker actions through /do. Use for automatic reactions to file changes; do NOT use for…
review
5-pass structured code review — correctness, security, performance, readability, consistency.
live-preview
Mid-build visual verification loop. Takes screenshots of components during construction, not just after. Catches visual regressions and invisible features before they compound. Requires Playwright or similar screenshot tool.
marshal
Meta-orchestrator that takes any direction — broad, specific, or vague — and autonomously chains skills and context into actionable work. Gathers context from codebase, docs, and memory. Only asks the user when it genuinely cannot proceed. Single-session orchestrator.
wiki
Markdown-first knowledge base where the LLM acts as librarian. Ingests raw sources, compiles and interlinks topic files, self-maintains an index. No vector DB or embeddings required -- uses LLM-native navigation over structured markdown up to 400K words.
finance-investor-relations
투자 유치를 위한 IR 피치덱과 3개년 재무 모델(매출 예측·손익·현금흐름·밸류에이션)을 만들어 드립니다. 다음과 같은 요청 시 사용하세요: "IR 덱 만들어줘" "투자자 피치 자료 써줘" "재무 모델 만들어줘" "매출 예측 좀 해줘" "밸류에이션 계산해줘" "시리즈 A 투자 유치 자료 준비해줘" "투자자 보고서 정리해줘" 투자 단계(Pre-SeedSeries B)에 맞춘 12슬라이드 피치덱·재무 모델·예상 질문 Q&A를 만들고, PPT/엑셀 변환으로 이어집니다.