agent-expert
Skill96 starsChanged 16 days agoMITView on GitHub
INTERNAL reference, NOT shipped with the installer. The in-repo expert on every coding-agent CLI CC Director can run (Claude Code, Codex, Gemini, OpenCode, Grok, Copilot, Cursor, Pi). Knows each agent's command line, config files, context-injection points, lifecycle hooks and events, SDK/API/server mode, transcript format, session semantics, and how CC Director drives it. Triggers on "agent-expert", "agent expert", "how does <agent> inject context", "<agent> hooks", "<agent> events", "<agent> CLI reference", "which agents support X", "agent capability matrix".
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 15
When you need to know, for any agent, how to:
- Distinctive: line 20
read its transcript, or drive it through an SDK / server / RPC channel,
- Very specific: line 50
Every agent reaches the fleet through its own Director (CC_DIRECTOR_API) and gets a launch preamble (its identity plus the cc-* fleet …
≈ “Common” is an estimate: at least 1 in 5 of the 308,232 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.
Files
Agent Expert (internal super-agent-manager)
This skill is the single place that knows everything we need to build drivers and plugins for the coding-agent CLIs CC Director supervises, and to make sessions talk to each other. It exists ONLY in this repository, for our own use. It is NOT deployed to customers and must be excluded from the installer skill bundle (see "Do not ship" below).
What it is for
When you need to know, for any agent, how to:
- launch it (command line, headless mode, session id, model flag, output format),
- configure it (config files and their precedence),
- inject context at launch (instruction files, system-prompt flags, hook additionalContext),
- get a signal when it clears or compacts its context (lifecycle hooks / event bus),
- read its transcript, or drive it through an SDK / server / RPC channel,
- and exactly how CC Director integrates it today plus the gaps -
read the per-agent reference file. Do not re-derive this from the web each time; that is the delay this skill removes. If a fact is missing or stale, research it, then update the file.
How to use
- Open
agents/README.mdfor the at-a-glance cross-agent matrix. - Open
agents/<agent>.mdfor the deep reference on one agent. These files are authoritative; the matrix is only the summary. - Each file marks every fact [VERIFIED from docs] or [INFERRED/UNCERTAIN] and links the source.
The agents
| Agent | AgentKind | File |
| Claude Code | ClaudeCode | agents/claude-code.md |
| OpenAI Codex | Codex | agents/codex.md |
| Google Gemini | Gemini | agents/gemini.md |
| opencode | OpenCode | agents/opencode.md |
| xAI Grok ("Grok Build") | Grok | agents/grok.md |
| GitHub Copilot | Copilot | agents/copilot.md |
| Cursor (cursor-agent) | Cursor | agents/cursor.md |
| pi | Pi | agents/pi.md |
To add an agent, copy agents/_template.md and fill every section.
The one thing this skill is really about: context injection and reset detection
Every agent reaches the fleet through its own Director (CC_DIRECTOR_API) and gets a launch
preamble (its identity plus the cc-* fleet commands). The hard part is keeping that preamble
present after the agent clears or compacts its context. The CLIs fall into four mechanism
families - see agents/README.md for which agent is in which:
- Family A - shell-command hook that injects context. A config file registers a command we
own; the command prints the preamble as additionalContext. This is what is already wired
for Claude (
ClaudeHookInstallerplus theGET /sessions/{sid}/fleet-preambleendpoint). - Family B - out-of-process event bus. We subscribe to the agent's event stream and push the preamble when we see a new/cleared/compacted session.
- Family C - in-process extension or plugin we author and load into the agent.
- Family D - instruction file the agent re-reads every prompt (passive, self-healing, no event).
The shared piece across all families is the Director endpoint GET /sessions/{sid}/fleet-preamble,
which already exists and is agent-agnostic.
Do not ship
This skill is internal. Before any installer/skill-bundle change, confirm agent-expert is
excluded from the shipped skill set. It documents competitor tooling and our internal
integration plans and has no place on a customer machine.
Other versions1 project uses a version of this.
Word for word · 1
- thefrederiksen/devthrottle★ 96100% same
Edited · 0
None found.
Wrote their own version · 0
None found.
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository30 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- cc-settings-apiSkill
- cencon-generateSkill
- code-maintenanceSkill
- commitSkill
- dead-code-finderSkill
- deploy-hosted-gatewaySkill
- developer-agentSkill
- dev-throttleSkill
- document-featuresSkill
- enter-issueSkill
- fleet-commsSkill
- implementation-loopSkill
- move-sessionSkill
- playwright-cliSkill
- product-agentSkill
- qa-agentSkill
- read-issuesSkill
- release-managerSkill
- review-codeSkill
- shipSkill
- skill-managerSkill
- support-agentSkill
- terminologySkill
- test-managerSkill
- ui-testSkill
- update-docsSkill
- voice-reviewSkill
- warning-fixerSkill
- wingman-brief-reviewerSkill
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
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.

