agentleFS
metagit-aimetagit-ai/metagit-cli/skills/metagit-control-center/SKILL.md

metagit-control-center

Skill4 starsChanged 13 days agoMITView on GitHub

Use when running metagit as an MCP control center for multi-repo awareness, guarded sync, and operational knowledge across ongoing agent tasks.

The file

Show its rules and text

We split this skill into the rules it gives an agent

18 rules. About 0 are common, 10 are distinctive, and 8 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈10

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 8

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this skill when an agent should actively coordinate repository context and task execution across a workspace.

    line 7
  2. Distinctive:

    Use when skill scripts are unavailable (Hermes SKILL-only install, no shell helpers):

    line 39
  3. Very specific:

    Run context pack + session prompt first (see metagit-context-pack skill):

    line 54
  4. Very specific:

    Read metagit://workspace/config and metagit://workspace/repos/status (MCP) or tier-1 pack JSON (CLI).

    line 62
  5. Very specific:

    Call metagit_project_context_switch when the objective is tied to a workspace project.

    line 63
  6. Very specific:

    Run metagit_workspace_health_check or read metagit://workspace/health for maintenance signals.

    line 64
  7. Distinctive:

    Use workspace search and upstream hints before broad exploration.

    line 69
  8. Very specific:

    Default to fetch for visibility.

    line 73
  9. Very specific:

    Use pull or clone only with explicit permission and rationale.

    line 74
  10. Very specific:

    Before switching projects: metagit_session_update (notes + recent repos), optional metagit_workspace_state_snapshot.

    line 78
  11. Very specific:

    After returning: metagit_workspace_state_restore when a snapshot was taken (metadata only.

    line 79
  12. Distinctive:

    When delegating single-repo or specialist work:

    line 89
  13. Distinctive:

    Use metagit search first when blocker appears external to current repo.

    line 107
  14. Distinctive:

    Prefer deterministic evidence over speculative jumps.

    line 108
  15. Distinctive:

    Keep operations minimal and auditable.

    line 109
  16. Distinctive:

    Never mutate repositories without explicit authorization.

    line 121
  17. Distinctive:

    Never broaden scope beyond configured workspace boundaries.

    line 122
  18. Distinctive:

    Always preserve a clear audit trail of control actions.

    line 123

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Metagit Control Center Skill

Use this skill when an agent should actively coordinate repository context and task execution across a workspace.

Purpose

Provide a repeatable control-center workflow where Metagit MCP guides awareness, synchronization, and operational continuity over multiple related repositories.

Bundled scripts (optional)

Helper scripts live under scripts/ when the full skill tree is installed. Hermes skill_manage copies SKILL.md only — scripts are not present unless you also run metagit skills install --skill metagit-control-center.

Resolve scripts from the PyPI package (works whenever metagit-cli is installed):

SKILL_ROOT="$(python3 -c "import metagit, pathlib; print(pathlib.Path(metagit.__file__).parent / 'data/skills/metagit-control-center')")"
"$SKILL_ROOT/scripts/control-cycle.sh" [root_path] ["query"] [preset]

Or from a full skill install:

"${HERMES_HOME:-$HOME/.hermes}/skills/metagit-control-center/scripts/control-cycle.sh" .

Wrapper behavior:

  • runs gating status first
  • optionally runs upstream discovery for blocker queries
  • emits compact, machine-readable lines

Inline CLI fallback (no scripts)

Use when skill scripts are unavailable (Hermes SKILL-only install, no shell helpers):

export METAGIT_AGENT_MODE=true
metagit mcp serve --status-once --root .
metagit context pack --tier 1 --json -c .metagit.yml
metagit prompt workspace -k health-preflight --text-only -c .metagit.yml
# blocker may be upstream — search managed repos:
metagit search "error signature" -c .metagit.yml --json
metagit prompt workspace -k sync-safe --text-only -c .metagit.yml

Core Workflows

1) Session Initialization

  • Run context pack + session prompt first (see metagit-context-pack skill):
metagit context pack --tier 2 --json -c .metagit.yml
metagit prompt workspace -k session-start --text-only -c .metagit.yml
  • Validate active workspace gate (metagit mcp serve --status-once or gate script).
  • Read metagit://workspace/config and metagit://workspace/repos/status (MCP) or tier-1 pack JSON (CLI).
  • Call metagit_project_context_switch when the objective is tied to a workspace project.
  • Run metagit_workspace_health_check or read metagit://workspace/health for maintenance signals.
  • Identify stale repos and unresolved blockers from prior activity.

2) Active Task Support

  • For each coding objective, map impacted repos.
  • Use workspace search and upstream hints before broad exploration.
  • Sync only repos that are required by the active objective.

3) Guarded Synchronization

  • Default to fetch for visibility.
  • Use pull or clone only with explicit permission and rationale.
  • Track sync outcomes in operations log resource.

4) Operational Memory

  • Before switching projects: metagit_session_update (notes + recent repos), optional metagit_workspace_state_snapshot.
  • After returning: metagit_workspace_state_restore when a snapshot was taken (metadata only; git tree is unchanged).

Maintain bounded local records of:

  • sync actions
  • issue signatures searched
  • candidate upstream repos identified
  • unresolved dependencies and follow-ups

5) Subagent dispatch loop

When delegating single-repo or specialist work:

  1. Route — metagit agent list --json or MCP metagit_agent_catalog; pick template_id from delegates_to.
  2. Plan — metagit agent dispatch-plan <id> --project P --repo R --vendor V --json or MCP metagit_agent_dispatch_plan.
  3. Profile — when JSON includes profile_apply_command, run it (or metagit agent apply …) before launch. <!-- modality:agent_profile_apply --> <!-- modality:dispatch_profile_capabilities -->
  4. Ensure — run install.command once when install.needed is true.
  5. Hand off — run handoff.context_pack and handoff.prompt; pass handoff.effective_instructions to the subagent.
  6. ACL isolation (optional) — when handoff.acl_commands is present, follow skill metagit-agent-coordination (allocate → lease → worktree → claim) so agents never share checkouts. <!-- modality:acl_branch --> <!-- modality:acl_lease --> <!-- modality:acl_worktree --> <!-- modality:acl_claim --> <!-- modality:acl_manifest --> ACL branch leases ≠ handoff claim TTL. Doc: docs/reference/agent-coordination.md.
  7. Reconcile — metagit_session_update after the subagent returns; respect out_of_scope boundaries.
metagit agent dispatch-plan repo-implementer \
  --project my-api --repo backend --vendor cursor --json

Decision Guidelines

  • Use metagit search first when blocker appears external to current repo.
  • Prefer deterministic evidence over speculative jumps.
  • Keep operations minimal and auditable.

Output Contract

For each control-center cycle, provide:

  • current objective
  • repositories examined
  • actions taken (or intentionally deferred)
  • next recommended step

Safety Rules

  • Never mutate repositories without explicit authorization.
  • Never broaden scope beyond configured workspace boundaries.
  • Always preserve a clear audit trail of control actions.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.