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/drubinstein/terry/sherlocknpx skills add drubinstein/terry --skill sherlockgit clone --depth 1 https://github.com/drubinstein/terryWrote 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/drubinstein/terry/sherlock)<a href="https://agentmods.dev/skills/drubinstein/terry/sherlock"><img src="https://agentmods.dev/badge/skills/drubinstein/terry/sherlock.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 | $0.00146 | $0.03323 |
| Opus 5 | $0.00073 | $0.01662 |
| Sonnet 5 | $0.00029 | $0.00665 |
| Haiku 4.5 | $0.00015 | $0.00332 |
Grade A, and why
sherlock 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 4d 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.
- (f) client / (h) harness: reproduced with a fresh `curl --no-keepalive` → ✗ both. How it starts
The opening of the file, as written. The whole thing — 219 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Sherlock — root-cause by elimination
"When you have eliminated the impossible, whatever remains, however improbable, must be the truth." — Sherlock Holmes, The Sign of the Four
Sherlock is the differential-diagnosis discipline for bugs. Instead of chasing your favorite theory, you build the complete list of possible causes, then eliminate each one with decisive evidence — never with a hunch or a reputation — until a single explanation survives. Whatever survives is the truth, however much it "shouldn't" be possible.
The whole method turns on one asymmetry: you cannot confirm your way to a root cause, but you can eliminate your way there. Confirmation flatters the theory you already hold; elimination is forced to consider the ones you don't.
Six moves, each a guard against the way bug-hunts actually fail:
- The complete suspect list. Enumerate all candidate causes up front — including the ones you "know" can't be it. An incomplete list is the #1 reason a bug "can't be found": the true cause was never written down.
- Eliminate by proof, not by reputation. A suspect leaves the board only when a test makes it impossible. "The stdlib is well-tested," "we decommissioned that," "the compiler doesn't have bugs" are not evidence — they're the exact assumptions the real bug hides behind.
- Falsify, don't confirm. Design the test that would kill your leading theory, not the one that flatters it. Twist theories to fit facts; never twist facts to fit theories.
- The dog that didn't bark. The absence of an expected signal — a missing log line, an error that never fired, a retry that never happened — is evidence. Observe, don't merely see.
- Trust the improbable survivor. When one cause remains, accept it even though it "can't" happen, then confirm it. The bug that can't happen is happening.
- Zero survivors ⇒ your reasoning failed, not the universe. If you've eliminated everything, you either over-eliminated (killed a suspect on assumption, not proof — re-test it) or under-enumerated (the cause is one you dismissed as impossible — widen the list). This branch is the maxim itself.
What ships with it
4 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.
- 4d ago First seen · 219 lines · 0 tokens per session scan A 05f4779b86d2
sherlock is a skill published in the GitHub repository drubinstein/terry (1 stars, last pushed 2mo ago), licensed MIT. It adds 146 tokens to every session and 3,323 once invoked, about $0.0007 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
autoimplement
Auto-advance a multi-phase plan: dispatch a subagent per phase, chain /review /pitfall-verification at each boundary, stop on actionable findings. Use for "autoimplement", "run this plan end-to-end", "auto-advance phases".
swiftui-design-consultation
Apple-canon design system for SwiftUI projects: produces DESIGN.md + a Swift Package starter (semantic colors, SF Pro, Liquid Glass, motion, accessibility). Use when starting or refreshing a SwiftUI design system.
ios-native-review
After a PRD/spec/plan for an iOS app, before implementation: validate the artifact against Apple HIG (iOS) via WebFetch citations. Asks "is this iOS-native?" — complements pitfall-verification and quality-review.
office-hours-track-aware
Track-aware wrapper for gstack /office-hours: runs the session, detects native vs web, renders + opens DESIGN.html before the approval gate. Use when starting a new product idea or brainstorming.
pitfall-verification
After any PRD, spec, plan, or code change: verify artifact-specific pitfalls don't apply before declaring done. For ship-worthy changes, auto-chains Codex a third model house + adversarial synthesis.
verify-and-land
Build the checked-out branch, launch that exact app, let the user confirm the fix is really there, then push and offer merge or PR.