Skip to content

cianfhoghlaim/cianchosaint / .agents/skills/agents-sync

agents-sync

Verified

Layer 10 of the knowledge-sync-loop — the agent definitions surface validator. Use when the user asks 'are all agents registered', 'is the agent fleet healthy', 'what are the NCCA subject specialists', 'does the agent registry match the 12-agent fleet', 'what does sync:agents do'. Per the 2026-08-15-agent-definitions-sync-loop-v1 change. Triggers: 'sync:agents', 'agent_definitions', 'agent-fleet-search', '12-agent fleet', 'NCCA subject specialists', 'agent_registry', 'agents:smoke', 'agents:audit', 'agents:reproduce'.

Agent Definitions Sync (Layer 10 of the knowledge-sync-loop)

The Layer 10 of the 10-layer pull-based sync architecture. Validates the 188 agent files + the 12-agent fleet + the 8 NCCA subject specialists.

Why Layer 10?

The 9-layer architecture covered 12 of the 14 knowledge surfaces. The next-biggest remaining gap was the agent definitions surface - the 188 .py agent files in 7 subdirs (agents/ + agents/adk/ + agents/agno/ + agents/api/ + agents/tools/ + agents/tuatha/ + agents/meaisinfhoghlaim/).

Agent definition drift is silent: when a developer adds a new agent to the 12-agent fleet but doesn't register it in agent_registry.py

  • doesn't update the routing keywords + doesn't add an AGENTS.md, the breakage doesn't surface until the next agent registration test. The 8 NCCA subject specialists (in agents/tuatha/agents/) are particularly prone to drift because they were added incrementally.

Layer 10 closes the agent definitions gap.

What Layer 10 covers

bash scripts/sync/agents.sh walks the 7 agent subdirs + produces a per-subdir report to stedding/sync-reports/agents-{date}.md with:

  • The per-subdir .py file counts
  • The 5 AGENTS.md files
  • The 8 NCCA subject specialists (in agents/tuatha/agents/)
  • The 12-agent fleet + the 4 NCCA subjects + the remaining subdirs
  • Drift detection (unregistered agents + stale model refs)

The orchestrator mise run sync:agents runs all 5 sub-layers + writes a unified report to stedding/sync-reports/agents-all-{date}.md.

The 5 sub-layers

Sub-layerTaskWhat it does
1sync:agents-driftDetects unregistered agents + stale model refs
2sync:agents-cccAppends the 25th CCC concept guide (agent-fleet-search) + reindex
3sync:agents-cogneeIngests the 188 agent files into the 14th Cognee cluster (agent_definitions)
4sync:agents-testRuns the agent registration test + reports pass/fail per agent
5sync:agents-lintPer-subdir stats (7 subdirs + 5 AGENTS.md + 8 NCCA subject specialists)

The new artifacts

ArtifactFilePurpose
agents-sync skill.agents/skills/agents-sync/SKILL.mdthis file
25th CCC guideagent-fleet-searchsurfaces the 188 agent files via CCC
14th Cognee clusteragent_definitionsthe 188 agent files
agents_sync_health assetorchestration/defs/sync_assets.pyDagster asset
scripts/cognee_ingest_agent_definitions.pyingestorThe canonical Cognee cluster ingestor
notebooks/sync_health.pydashboardAgents tab in the grouped sync-health marimo surface

Agent definitions evolution feedback loop

The system grows its knowledge surface over time via the agent definitions evolution feedback loop:

agent file modified
  → sync:agents-cognee detects the change
  → re-cognifies the modified agent into the agent_definitions cluster
  → sync:agents-ccc updates the 25th concept guide
  → The deployment control panel (notebook 24) surfaces the change

Quick routing

If you want to...Do this
Check the agent fleet healthmise run sync:agents
See the per-subdir breakdowncat stedding/sync-reports/agents-$(date +%Y-%m-%d).md
Add a new agentCreate the .py file in the right subdir + register it in agent_registry.py + add it to routing keywords + update agents/AGENTS.md + run sync:agents
Fix a registration driftsync:agents-drift will list the unregistered agents; register + re-run
See the agent definitions dashboardOpen notebooks/sync_health.py (Agents tab)
Run the agent registration testmise run agents:smoke

Cross-references

  • openspec/changes/2026-08-15-knowledge-sync-loop-v1/ (the foundation)
  • openspec/changes/2026-08-15-stacks-sync-loop-v1/ (Layer 8)
  • openspec/changes/2026-08-15-dlt-sync-loop-v1/ (Layer 9)
  • agents/ (the 12-agent fleet + the 188 .py files)
  • agents/AGENTS.md (the canonical agent documentation)
  • agents/tuatha/ (the 8 NCCA subject specialists)
rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit.Stars: 469kUpdated Claude CodeCodexVerifiedInstall
rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure.Stars: 469kUpdated Claude CodeCodexVerifiedInstall
rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics.Stars: 469kUpdated Claude CodeCodexVerifiedInstall
ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first.Stars: 194kUpdated Claude CodeCodexVerifiedInstall