Turn Claude Code into a team that has to prove it. 44 skills, 8 review agents, a validation router and a knowledge layer where nothing is claimed without a check that could have failed.
Set up Claude Code hooks to block dangerous git commands (push, reset --hard, clean, branch -D, etc.) before they execute. Use when user wants to prevent destructive git operations, add git safety hooks, or block git push/reset in Claude Code.
Break a plan, spec, or the current conversation into a set of tracer-bullet tickets, each declaring its blocking edges, published to the configured tracker (edges as text in one file per ticket locally, or native blocking links on a real tracker).
Build a throwaway build-a-throwaway-prototype to answer a design question. Use when the user wants to sanity-check whether a state model or logic feels right, or explore what a UI should look like.
Use when deciding whether a table, a file, or a service can be deleted because it looks unused. Measure real access, rename before deleting, and check the units of every counter.
Use before pulling someone else's code, skills, prompts or configuration into your project, and before publishing anything that contains them. Turns "it was on GitHub" into an obligation you can actually meet.
Use when deciding how much review a change needs, and of what kind. What the change touches decides the level — not how big it looks, not how confident you feel.
Use when something did not happen and nothing complained — no error, no alert, an empty log, a change with no effect. Teaches how to read absence as a symptom and where silence gets manufactured.
Shared vocabulary for designing deep modules. Use when the user wants to design or improve a module's interface, find deepening opportunities, decide where a seam goes, make code more testable or AI-navigable, or when another skill needs the deep-module vocabulary.
Use when checking whether curated knowledge is still fresh, reading freshness verdicts, or investigating why a unit shows drifted or unknown. Triggers on requests like "check freshness", "run the probes", "is this knowledge still current", "why does kb-0002 say unknown", or "read the knowledge index".
Diagnosis loop for hard bugs and performance regressions. Use when the user says "diagnose"/"debug this", or reports something broken/throwing/failing/slow.
Use when recording a new piece of curated knowledge -- a finding, a decision, a measured fact worth keeping and later re-checking for freshness. Triggers on requests like "record this as knowledge", "write a knowledge unit for X", "capture this finding with evidence", or "add this to curated knowledge". Do not use for…
Generate an interactive bash generate-a-setup-wizard that walks a human through steps only they can perform. Use when provisioning infrastructure, setting up credentials or CI secrets, walking an unfamiliar third-party dashboard, or running a one-off migration or cutover. Don't invoke this for steps the agent can…
Use when a claim needs a second opinion that does not share your blind spots. How to ask a different model, what to send it, and how to settle a disagreement.
Use when entering a system you did not build. The reading order that gets you oriented without breaking anything, and the false negatives that will fool you on the way.
Use when recording or updating a persistent agent-memory fact -- a user preference, a project fact, feedback, or a reference note the harness should remember across sessions. Triggers on requests like "remember that I prefer X", "note this project fact", "update this memory", or "this preference changed". Do not use…
Use before committing configuration, when a secret has already been committed, or when adding a secret scanner. Covers why rotation beats scrubbing and why most scanners get switched off.
Use before shipping something that runs repeatedly — a job, a query, an agent loop, a pipeline. Estimates the recurring cost of a change while it is still cheap to change, and decides where spending less is safe.
Use after an incident, or before shipping something whose failure would be invisible. Decides what to log, measure and alert on so the next failure announces itself instead of being discovered by a customer.
Use when documenting a system or auditing what is already documented. Records the edge of what is known, so that silence stops reading as "nothing there".