Collider.club
CPO at Collider
Architecture decision record explaining why the collection ships as a Claude Code plugin and why no Codex plugin exists yet. Useful context for anyone packaging their own skills for more than one harness.
Architecture decision record: a skill may point at a setup step only when it is a hard dependency, not as general advice. Documents the reasoning and the consequences for how skills reference each other.
The canonical install instructions, copied verbatim wherever installation is documented so the steps never drift between README and docs. Covers the Claude Code plugin route and the marketplace fallback.
Template and procedure for the human-facing docs page every promoted skill carries. Fixes the four required sections — what it does, when to reach for it, common questions, it's working if — plus section order and where to source the questions.
Defines the two invocation modes every SKILL.md must declare: user-invoked, reachable only by a human via slash command, or model-invoked, which the agent may trigger on its own. Explains the frontmatter and the openai.yaml policy flags that enforce each.
The router skill: maps every user-reachable skill and how they relate, so an agent can pick the right one instead of guessing. Kept in sync whenever a skill is added, renamed or removed, on the principle that a router which lies is worse than none.
Shared domain vocabulary for the repository. Defines Issue tracker, Issue, Decision ticket and Triage role, lists the terms to avoid, maps how the entities relate and records resolved ambiguities, so agent and human use one language instead of drifting synonyms.
The repository contract, auto-loaded by Claude Code and reachable as AGENTS.md through a symlink so Codex-style agents read the same file. Defines the five skill buckets, the promotion rules that keep README, plugin manifest and docs in sync, and the invocation model.
Index of the productivity bucket: seven skills for non-code workflow, each linked to its SKILL.md. Covers grill-me, grilling, handoff, teach, to-questionnaire, wait-what and writing-for-agents.
Index of the engineering bucket: eighteen skills for daily code work, each with a one-line description linked to its SKILL.md. Includes tdd, code-review, diagnosing-bugs, codebase-design, domain-modeling, implement, research, triage and wayfinder.
Full release history of the skills collection, maintained with changesets. Records every skill added, renamed, promoted between buckets or deprecated, which is the fastest way to see how the set has evolved and what is currently moving.
Front door of the pack: what these skills are, how to install them into Claude Code or another harness, and a linked index of all promoted skills split into user-invoked and model-invoked. Start here to see what the collection actually contains.
Navigation across mattpocock/skills with measured metrics: 101 markdown files out of 151, 88% of the repo by bytes, 35 SKILL.md across five buckets. Covers the promotion contract, the user-invoked vs model-invoked split, the AGENTS.md symlink trick and a single MIT license zone.
Bootstrap file for AI agents only; human readers are redirected to README.md. Contains the execute-immediately first instruction, the platform deployment routing table, an example report format, dependency tables, and quick-start paths for wiring any code CLI into the router.
Root-level Kali Linux shortcut entry. Defers to kali/README-kali.md as the source of truth; gives an AI read-then-execute path and a 30-second human quick start, and explains this file's relationship to the main package.
Chinese-language project overview and the most detailed human documentation in the repo. Mirrors README.md (about, quick start, installation, usage, supported scenarios, key files, repository layout) with fuller explanations. Primary reference for Chinese-speaking users.
Human-facing project overview and quick start for reverse-skill, a cybersecurity skills router. Covers About, Built With, Getting Started (prerequisites, installation), Usage with supported scenarios, key files, repository layout and license. The entry point for human contributors.
Chinese edition of the routing matrix (by target type, by user intent, route-not-matched handling, cross-module path crossing). Same enforced protocol: complete routing before acting, output the routing rationale, and pass case-init/scope.md before ACT on a target.
Full task-to-skill routing matrix. Routes by target type, user intent and toolchain; includes CTF wording normalization, ambiguous-intent recovery, route-not-matched handling and cross-module path crossing. Enforced, not advisory: routing must complete before execution.
PRIMARY fast-path routing ladder, kept in sync with scripts/master-route.ps1. Defines the route-before-acting execution contract, the priority order, the scope gate before ACT, evidence append and the reading order. The first file an agent consults.
Collider.club
CPO at Collider
Architecture decision record explaining why the collection ships as a Claude Code plugin and why no Codex plugin exists yet. Useful context for anyone packaging their own skills for more than one harness.
Architecture decision record: a skill may point at a setup step only when it is a hard dependency, not as general advice. Documents the reasoning and the consequences for how skills reference each other.
The canonical install instructions, copied verbatim wherever installation is documented so the steps never drift between README and docs. Covers the Claude Code plugin route and the marketplace fallback.
Template and procedure for the human-facing docs page every promoted skill carries. Fixes the four required sections — what it does, when to reach for it, common questions, it's working if — plus section order and where to source the questions.
Defines the two invocation modes every SKILL.md must declare: user-invoked, reachable only by a human via slash command, or model-invoked, which the agent may trigger on its own. Explains the frontmatter and the openai.yaml policy flags that enforce each.
The router skill: maps every user-reachable skill and how they relate, so an agent can pick the right one instead of guessing. Kept in sync whenever a skill is added, renamed or removed, on the principle that a router which lies is worse than none.
Shared domain vocabulary for the repository. Defines Issue tracker, Issue, Decision ticket and Triage role, lists the terms to avoid, maps how the entities relate and records resolved ambiguities, so agent and human use one language instead of drifting synonyms.
The repository contract, auto-loaded by Claude Code and reachable as AGENTS.md through a symlink so Codex-style agents read the same file. Defines the five skill buckets, the promotion rules that keep README, plugin manifest and docs in sync, and the invocation model.
Index of the productivity bucket: seven skills for non-code workflow, each linked to its SKILL.md. Covers grill-me, grilling, handoff, teach, to-questionnaire, wait-what and writing-for-agents.
Index of the engineering bucket: eighteen skills for daily code work, each with a one-line description linked to its SKILL.md. Includes tdd, code-review, diagnosing-bugs, codebase-design, domain-modeling, implement, research, triage and wayfinder.
Full release history of the skills collection, maintained with changesets. Records every skill added, renamed, promoted between buckets or deprecated, which is the fastest way to see how the set has evolved and what is currently moving.
Front door of the pack: what these skills are, how to install them into Claude Code or another harness, and a linked index of all promoted skills split into user-invoked and model-invoked. Start here to see what the collection actually contains.
Navigation across mattpocock/skills with measured metrics: 101 markdown files out of 151, 88% of the repo by bytes, 35 SKILL.md across five buckets. Covers the promotion contract, the user-invoked vs model-invoked split, the AGENTS.md symlink trick and a single MIT license zone.
Bootstrap file for AI agents only; human readers are redirected to README.md. Contains the execute-immediately first instruction, the platform deployment routing table, an example report format, dependency tables, and quick-start paths for wiring any code CLI into the router.
Root-level Kali Linux shortcut entry. Defers to kali/README-kali.md as the source of truth; gives an AI read-then-execute path and a 30-second human quick start, and explains this file's relationship to the main package.
Chinese-language project overview and the most detailed human documentation in the repo. Mirrors README.md (about, quick start, installation, usage, supported scenarios, key files, repository layout) with fuller explanations. Primary reference for Chinese-speaking users.
Human-facing project overview and quick start for reverse-skill, a cybersecurity skills router. Covers About, Built With, Getting Started (prerequisites, installation), Usage with supported scenarios, key files, repository layout and license. The entry point for human contributors.
Chinese edition of the routing matrix (by target type, by user intent, route-not-matched handling, cross-module path crossing). Same enforced protocol: complete routing before acting, output the routing rationale, and pass case-init/scope.md before ACT on a target.
Full task-to-skill routing matrix. Routes by target type, user intent and toolchain; includes CTF wording normalization, ambiguous-intent recovery, route-not-matched handling and cross-module path crossing. Enforced, not advisory: routing must complete before execution.
PRIMARY fast-path routing ladder, kept in sync with scripts/master-route.ps1. Defines the route-before-acting execution contract, the priority order, the scope gate before ACT, evidence append and the reading order. The first file an agent consults.