AI & Agents Agent Skills

Run gitnexus analysis for a target workspace and selected project repositories before graph-dependent tasks. Use when index staleness is detected or cross-repo graph results are needed.

Discover, suggest, promote, validate, and ingest cross-repo graph.relationships in .metagit.yml for GitNexus overlay and dependency maps. Use for first-time graph bootstrap or ongoing workspace graph maintenance.

Token-efficient MCP resource ladder for metagit workspaces — catalog, map, layered prompts, project/repo drill-down. Use when the IDE host exposes metagit resources/read instead of shell CLI.

Ongoing workspace and project management for OpenClaw and Hermes agents. Use when starting work, organizing repos, or before creating a new project folder so existing metagit projects are reused instead of duplicated.

Orchestrate a reference-implementation rewrite across source and target repos using campaigns, parity registry conventions, objectives, and subagent handoffs.

Configure shared coordination state (objectives, handoffs, approvals, events) across multiple agents and machines via METAGIT_STATE_URL and the ops HTTP backend. Use when Hermes subagents, CI runners, or humans must see the same objective queue without Syncthing JSON files.

Discover active metagit workspace scope, project boundaries, and repository status. Use when an agent starts work in a multi-repo workspace and needs fast, scoped context before editing.

Sync workspace repositories safely using metagit with scoped fetch, pull, or clone actions. Use when repository content must be refreshed for implementation.

Optimize any repository for minimal-token agent onboarding — llms.txt, AGENTS.md, hidden HTML agent blocks, docs/agents.md. Use when asked to improve agentic access, agent metadata, llms.txt, or run the agent-access optimizer subagent on a project.

Isolate concurrent agents with RFC-0007 ACL primitives: allocate agent/* branches, acquire branch leases, create exclusive git worktrees, and declare advisory file claims. Use when launching multiple agents, avoiding shared checkouts, preventing branch collisions, or coordinating file ownership before coding. Distinct from handoff claim TTL leases.

Compose ACL, task graph, scheduler, merge, and related coordination subsystems via metagit aos|coord status|doctor|next. Use when an operator or orchestrator needs one snapshot of coordination health or the next work envelope without launching models.

Use when generating or refining local .metagit.yml files using deterministic discovery plus MCP sampling or CLI-only fallbacks.

CLI-only shortcuts for metagit agents — workspace catalog, discovery, prompts, sync, layout, and config. Use instead of MCP or HTTP API when operating from a shell or agent_mode session.

CLI-only shortcuts for metagit agents — workspace catalog, discovery, prompts, sync, layout, and config. Use instead of MCP or HTTP API when operating from a shell or agent_mode session.

Run gitnexus analysis for a target workspace and selected project repositories before graph-dependent tasks. Use when index staleness is detected or cross-repo graph results are needed.

Discover, suggest, promote, validate, and ingest cross-repo graph.relationships in .metagit.yml for GitNexus overlay and dependency maps. Use for first-time graph bootstrap or ongoing workspace graph maintenance.

Inspect PR failed checks and unresolved review comments, fix valid feedback, push, and safely handle publish-via-PR workflows. Use when user says /pr-triage, asks about PR status, CI failures, review comments, or wants changes published through a PR.

Offline X audience simulation and evaluation. Use when evaluating draft tweets, handle bios, pinned-post ideas, or promotion angles against real scraped X context without posting or mutating X state.

Inspect PR failed checks and unresolved review comments, fix valid feedback, push, and safely handle publish-via-PR workflows. Use when user says /pr-triage, asks about PR status, CI failures, review comments, or wants changes published through a PR.

Offline X audience simulation and evaluation. Use when evaluating draft tweets, handle bios, pinned-post ideas, or promotion angles against real scraped X context without posting or mutating X state.

Reproduce and verify the benchmark claims of a tool or paper already triaged in the research repo. Locates the benchmark harness, runs it, and records verified vs. reported figures in benchmarks/sources/{slug}-repro.md. Triggers: reproduce benchmark, verify claims, run benchmark, check benchmark, benchmark reproduction, verify numbers.

Capture cross-project connections on the fly. Use when you notice a shared pattern, people-bridge, knowledge cascade, narrative amplifier, or flywheel loop linking two projects. Persists structured YAML to a bridges directory. Triggers on: bridge, connects to, feeds into, same pattern as, cross-project link, archetype detection, bidirectional, one-way, active, potential, theoretical.

Close out a pg-claude session — sync progress/STATE.md, progress/coverage.md, progress/files-examined.md, and append a sessions/ log entry whenever a session produced durable output (a new knowledge/idioms or knowledge/subsystems doc, a [verified-by-code] fact, a discovered gotcha, a file-by-file deep read, or a locked decision from pg-claude-plan.md §14). Use proactively when the user says "wrap up", "close out the session", "sync memory files", "record this gotcha", "we're done for the day", or "log this for next time" — and whenever durable output was just produced. Skip for PG MemoryContext / palloc / pfree internals questions (use memory-contexts), LangChain / LangGraph / LlamaIndex agent memory and vector stores, application memory leaks / valgrind / heaptrack, conversational chat history persistence, finding old Claude sessions (use find-session), and ChatGPT / Claude.ai conversation export.

Use only on an exact Harness activation, either the command $harness or the skill link [$harness:harness](.../skills/harness/SKILL.md). Runs the Harness workflow loop with classification, Plan approval, role-separated Implement, clean-context Review, Repair Plan approval, and Completion gates. Do not activate from documentation, quotations, escaped text, ordinary requests, or implicit inference.