armory

Plugin armory from marketplace agent-armory in nafania217518/agent-armory. Commands read from the repo's manifest — yesterday.

Skills (12)

• audit

  Quality audit of a whole repo or package: bugs, security holes, what breaks under real load, risky code without tests, slow paths, what to delete, and the tech-debt ledger. One-shot ranked report, changes nothing. Use for "audit this codebase", "review the whole repo", "find bloat", "what can I dele

• code

  Runs a coding task from the approved plan to the merge or PR: task branch, coder briefs, review and fix loop, task report, finish question. Load it before the first code search or edit of any coding task: "/code", "implement", "fix", "refactor", "add a feature".

• core

  Always-on base rules for the main session and every subagent; the hook injects them at session and subagent start. Use for "/core" and "show the armory base rules".

• debug

  Diagnosis of a bug whose cause is not known: feedback loop, reproduce and minimise, ranked hypotheses, one-variable probes, tagged debug logs, cleanup, redaction. Loaded by the debugger card, and by `code` when the main session debugs itself; "/debug".

• delegate

  How the main session spawns project agents: the brief, the worktree, the model and effort per spawn, and retries. Load it before any spawn or hand-off to an agent: "delegate", "spawn an agent", "which model", "write a brief", "/delegate".

• implement

  The coder's rules: scope before you write, the smallest complete change, readability, domain types and boundaries, root-cause fixes, safe refactors, the test-scope list, the gates per commit, atomic commits. Loaded by the coder and debugger cards and by `code`; "/implement". For a coding request fro

• plan

  Plans work in plan mode before any edit, at three levels: strategic (design doc and milestones), milestone (scope and BR requirements) and task (TR requirements and the plan file), with quizzes and the user's approval. Use for a new task, "plan this", "make a plan", "plan a milestone", "design doc",

• research

  How to research a technical claim: official sources first, independent sources next, dates, disagreements, the result format and the report slots. Load for any fact about a library, API, model, tool or domain that may be outdated, for the plan's research step, and in the researcher agent. Use for "c

• retro

  Retrospective on the current session: ranked findings on what to change in the project or in armory so that later sessions go better. Changes no file. Use for "/retro", "run a retro".

• review

  Review of a change: logic, safety, real load, tests for risky code, gates, lean and readable lines. Reads the connected code, not only the diff. Lists findings, changes no code. Use for "review this", "review the last commit", "review my PR", "is this over-engineered", "/review".

• setup

  Sets up armory in the current project: probes the hook, asks the user what to install, then writes the project context and the agent cards. Use for "/setup", "set up armory", "add the project agents".

• triage

  Triage of inbound tracker items on a bound tracker: the attention list, one item at a time, a recommended category and outcome, then the writes the user approves. Changes no repo file. Use for "/triage", "triage the tracker".

Hooks (2)

• SessionStart

  "${CLAUDE_PLUGIN_ROOT}/scripts/core-context.sh"

• SubagentStart

  "${CLAUDE_PLUGIN_ROOT}/scripts/core-context.sh"

Similar plugins