add-agent
Skill71 starsChanged 36 days agoMITView on GitHub
Add or substantially change an analysis agent, graph node, or LLM seat in this repository. Use when introducing a new agent role or changing its prompt, tools, state, routing, persistence, attribution, and tests.
9 other projects use a version of this.
Adopted by
Most recently taken up by
- venkateshv1266/pi-harnesstheir own version, first seen Nov 1, 2026
- chacosoldier/compabobtheir own version, first seen Oct 23, 2026
- axrmxv/claude-configtheir own version, first seen Oct 13, 2026
- FISHers6/clawBrotheir own version, first seen Oct 8, 2026
- mattolson/agent-sandboxtheir own version, first seen Oct 8, 2026
- microsoft/Sicotheir own version, first seen Jun 11, 2026
Used with
Projects with it also use setup.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
VercelSkillsetup
Set up Vercel CLI and project configuration. Use when the user says "set up Vercel", "configure Vercel", "link to Vercel", or "vercel init".
Also in 3 projects that use this
Other versions9 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 9
- microsoft/Sico★ 253same name
- mattolson/agent-sandbox★ 212same name
- BerriAI/litellm-skills★ 84same name
- chacosoldier/compabob★ 34same name
- FISHers6/clawBro★ 28same name
- venkateshv1266/pi-harness★ 2same name
- axrmxv/claude-config★ 1same name
- theopengroup/rune-harness★ 1same name
- tjg37/ironclaw★ 1same name
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 repository11 other files it gives its agents.
- AGENTS.mdAGENTS.md
- analyze-tickerSkill
- diagnose-runSkill
- maintain-agent-metadataSkill
- run-pipelineSkill
- run-testsSkill
- add-agentSkill
- analyze-tickerSkill
- diagnose-runSkill
- run-testsSkill
- validation-batchSkill
The file
We split this skill into the rules it gives an agent
8 rules. 0 are common, 4 are distinctive, and 4 are very specific to this project.
0
Common
Most of the 7 edited versions we read keep these.
4
Distinctive
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 14
Read docs/AGENT_ROSTER.md, docs/PROMPT_CONTRACTS.md, and the relevant graph and runtime modules.
- Very specific: line 16
Add or update the prompt through structured JSON handling under prompts/.
- Distinctive: line 23
Do not introduce a context-name exception or construction-time flex cap.
- Very specific: line 24
Add node logic in the owning src/agents/ module and tools in the owning src/tools/ module.
- Distinctive: line 26
Update state, reducers, graph wiring, retry/fallback behavior, output budgets, persistence, attribution, and publication requirements only …
- Distinctive: line 29
Add contract tests for seat registration, prompt loading, graph topology, attribution, malformed/refused output, retry behavior, and …
- Very specific: line 34
Run make test-prompts, the relevant targeted tests, and make check-all.
- Distinctive: line 36
Do not copy thresholds, model defaults, or enum values into this skill.
“Common” is measured: kept by at least 6 in 10 of the 7 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Add an analysis agent
First determine whether the requested concept is an LLM seat, deterministic validator, tool, or orchestration node. Extend the owning abstraction instead of forcing every concept into an agent shape.
For an LLM-backed role:
- Read
docs/AGENT_ROSTER.md,docs/PROMPT_CONTRACTS.md, and the relevant graph and runtime modules. - Add or update the prompt through structured JSON handling under
prompts/. - Register the canonical seat and capabilities in
src/llm_runtime/seats.py; construct it throughsrc/llm_runtime/construction.py. Choose its timeout class fromsrc/llm_runtime/timeouts.py. The runtime owns total application attempts; transports configure one SDK attempt per dispatch, with no per-seat retry setting. This does not prove one physical send: Google's aiohttp connection recovery can resend internally. Verify application retry and rate-limit branch counts. Do not introduce a context-name exception or construction-time flex cap. - Add node logic in the owning
src/agents/module and tools in the owningsrc/tools/module. Tool execution must stay inside the inspection and audit chain. - Update state, reducers, graph wiring, retry/fallback behavior, output budgets, persistence, attribution, and publication requirements only where the role needs them. Parallel branches require isolated fields or an intentional reducer.
- Add contract tests for seat registration, prompt loading, graph topology, attribution, malformed/refused output, retry behavior, and saved-artifact semantics. Exercise the actual adapter's dispatch options under an exhausted or nearly spent deadline, including corrections, degraded-tier dispatch and failover. A model's configuration fields alone do not prove its request deadline or retry options.
- Run
make test-prompts, the relevant targeted tests, andmake check-all.
Do not copy thresholds, model defaults, or enum values into this skill. Read them from the owning code and update that source plus its tests when behavior changes.
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.

