Establishes the operator identity, access authority, and execution mandate for all agent sessions. Always load at session start. The agent is the full operator and executes responsibilities through single-thread baton handoff (Manager -> Collaborator -> Admin -> Consultant). The user is the client, consulted only for…
Apply a low-resource operating profile for Playwright MCP + Claude Vision on constrained machines. Use this at session start, before visual QA, or after browser-context/OOM failures. Establishes bounded retries, local-vs-remote execution split, and optimized screenshot/vision practices.
Validate and enforce release version integrity across tags, manifests, changelogs, and publish workflows. Use before tagging, publishing, or when release metadata drift is suspected.
Onboard any repository into a standardized Copilot + CI governance baseline. Use for new repos and first-time sessions to ensure immediate standards adoption.
Audit and harden repository profile, community health, discoverability metadata, and contribution surfaces across repos using bounded, evidence-based checks.
Adversarial cross-family review of baton artifacts, PRs, and governance docs. Dispatches a non-Anthropic fleet model (cross-family invariant) and returns structured findings with ACCEPT/REJECT/PARTIAL verdicts and hallucination-risk classifications. Operates as a bounded, non-blocking advisory layer.
Prevent secret leakage across git history, package artifacts, logs, and docs. Use when editing workflows, packaging configuration, environment files, or release automation.
Run a bounded self-annealing review of operating instructions and workflow outcomes. Use this after failures, after long sessions, before merge, or when repeated mismatches appear between intended process and observed execution.
★not rated 1 1mo agoA46 tokens
originalMIT
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: