agents-remember

Skills Reference

Installed skills live under:

ar-coordination/skills/

The canonical source copies live under:

agents-remember/skills/

After editing root skills/, run:

python3 scripts/sync-skills.py

That refreshes the generated MCP package-data copy and every harness starter package skill folder. The pre-push hook runs python3 scripts/sync-skills.py --check so unsynced copies fail before push.

Lifecycle And Workflow Skills

Skill Purpose
l-01-session-job-lifecycle The session job lifecycle the coordinator routes every session into (orient → ground → frame → decide → build → close); owns the read-only exit and the build-mode decision.
w-02-light-task-workflow Durable one-page task plan with approval gate and live checklist; escalates to a master + light sub-task series for larger work.

Core Skills

Skill Purpose
c-00-initialize-memory-repo Create or repair a target repository memory root.
c-01-findings-capture Route durable findings to the right artifact.
c-02-memory-quality-control Control memory quality with task-start drift checks, pre-code-commit missing-onboarding checks, and closeout quality gates.
c-03-repo-bootstrap Bootstrap repo overviews, route-local overviews, and onboarding coverage.
c-04-retrieval-strategy-router Choose Semantics, Relationship, or Intent retrieval, using providers for discovery and onboarding/source as proof.
c-05-create-or-update-onboarding-files Create and maintain file-level onboarding and entity catalogs.
c-08-ar-coordination-context-resolver Resolve memory, coordination, task, temp, and cross-repo facts.
c-09-git-worktree-manager Manage worktree lifecycle, integration, and cleanup gates.
c-10-adopt-memory-baseline Adopt existing external-memory onboarding into the first ledgered baseline.
c-11-memory-carryover-from-branch Carry richer memory forward after matching code lands.
c-12-closeout Own the closeout approval gate and the code → memory → ledger commit sequence, for both direct edits and worktree-backed tasks.
c-13-install-and-onboard Lead package-based first-run setup after the harness starter package and MCP server are wired: preflight checks, runtime_install, memory repo, onboarding bootstrap, and provider indexing.

Installing Skills Into Harnesses

The normal first-run path does not use this section. Copy the harness starter package instead; it already contains the skill folders each harness discovers.

For manual maintenance or non-package installs, use:

skills_install()

The install target is normally inferred from the MCP settings location: <registration-root>/mcp/<settings>.json installs into <registration-root>/skills/. Use harnessSkillRoot only for non-standard harness layouts.

The packaged skills are flat (one folder per skill), so each installs as <registration-root>/skills/<skill-name>/ named by its lowercase frontmatter name.

Do not copy individual skill folders by hand for a maintenance install. The MCP tool copies the packaged skill tree consistently, including sibling files and shared helper modules.