Individual skill
octocode-subagent
Use when spawning workers or offloading work: Task/subagents, specialist handoffs, A2A peers, sealed packets, coordinate/synthesize, challenge techniques (rubber-duck, interview, mimic-flow, red-team, blind review, consensus), or local Ollama one-shots to save tokens. Measuring keep/discard KPIs → octocode-graph-eval.
skills/octocode-subagent/SKILL.md
Source commit c265e3f9413161c900cbf4aa70d451b8e6b3920a
What was checked
- Matches the observed repository HEAD
- Yes
- Scanner rules are current
- No
- Repository scan complete
- No
- Skill scan complete
- Yes
- Observed behavior
- Not checked yet
- Last checked
- Nov 4, 2026
These are observations at the recorded commit, not a guarantee of safety. Review the source before running it.
Install this skill
Choose an agent in your Stack or copy a command below. Commands fetch the pinned commit and copy only this skill directory. Existing destinations are refused. Copying a command does not install anything.
An install command is unavailable: current rules, complete repository and skill scans, an eligible behavior level and a permitted license are required.
Repository context
- Repository stars (not skill ratings)
- 957
- Repository last updated
- Nov 6, 2026
- License
- MIT
Observed agent integration: No specific integration evidence. Standard SKILL.md installation does not imply verified compatibility.
- ai Repository topic
- claude-ai Repository topic
- cursor-ai Repository topic
- development Repository topic
- mcp Repository topic
- modelcontextprotocol Repository topic
- octocode Repository topic
- ai-agents Repository topic
- ai-tools Repository topic
- code-intelligence Repository topic
- engineering Derived category
Files and excerpt
File metadata and permitted, redacted excerpts come from the recorded scan. Source links are pinned to that commit.
Browse files
README.mddoc · 2,143 bytesSKILL.mdcore · 6,210 bytesreferences/a2a.mddoc · 1,312 bytesreferences/blind-review.mddoc · 1,577 bytesreferences/consensus.mddoc · 1,764 bytesreferences/coordinate.mddoc · 1,358 bytesreferences/decision-matrix.mddoc · 2,238 bytesreferences/decompose.mddoc · 1,274 bytesreferences/family-playbooks.mddoc · 3,675 bytesreferences/improve-loop.mddoc · 569 bytesreferences/interview.mddoc · 1,956 bytesreferences/local-ollama.mddoc · 2,963 bytesreferences/mimic-flow.mddoc · 2,279 bytesreferences/model-routing.mddoc · 1,274 bytesreferences/model-selection.mddoc · 4,066 bytesreferences/octocode.mddoc · 676 bytesreferences/ollama-cli-run.mddoc · 2,029 bytesreferences/ollama-cli.mddoc · 3,144 bytesreferences/ollama-invoke.mddoc · 3,161 bytesreferences/ollama-local-models-heavy.mddoc · 1,971 bytesreferences/ollama-local-models-kits.mddoc · 2,132 bytesreferences/ollama-local-models-matrix.mddoc · 4,279 bytesreferences/ollama-local-models-sources.mddoc · 3,334 bytesreferences/ollama-local-models.mddoc · 4,004 bytesreferences/output.mddoc · 1,132 bytesreferences/packet-contract.mddoc · 1,915 bytesreferences/packet-schemas.mddoc · 1,307 bytesreferences/packets.mddoc · 1,623 bytesreferences/patterns.mddoc · 2,544 bytesreferences/recovery.mddoc · 1,353 bytesreferences/red-team.mddoc · 1,763 bytesreferences/references.mddoc · 2,755 bytesreferences/rubber-duck.mddoc · 1,679 bytesreferences/spawn-gate.mddoc · 1,406 bytesreferences/synthesize.mddoc · 1,582 bytesreferences/techniques.mddoc · 2,517 bytesreferences/usage-matrix.mddoc · 4,355 bytesreferences/verify-gate.mddoc · 2,285 bytesreferences/workflow.mddoc · 3,040 bytesreferences/workspace.mddoc · 697 bytesscripts/ollama-health.shruns · 2,187 bytesscripts/ollama-worker.shruns · 9,155 bytes
Read excerpt
# Octocode Subagent
Host-agnostic delegation: cloud/host workers **or** local Ollama offload.
Flows: `GATE → DECOMPOSE → ROUTE → PACKET → SPAWN → COORDINATE → SYNTHESIZE → CLEANUP` (tool-using) · `GATE → ROUTE → RUN → VERIFY → REPORT` (Ollama — `references/local-ollama.md`).
## Lobby rules
1. Spawn only when delegation changes speed, expertise, isolation, or context quality; otherwise keep work in the parent. Default is solo; earn spawn cost first.
2. One bounded objective per worker; no nested spawning unless the host explicitly allows it.
3. Workers inherit no parent chat: every packet carries goal, scope, context, authority, constraints, evidence needs, and return shape.
4. Treat worker output as claims; re-check load-bearing anchors (Ollama: always VERIFY).
5. Barrier before synthesize — wait/list every live worker (or stop+remove); merge conflicts first; then answer.
6. Parent owns the user, synthesis, and mutations unless a packet explicitly transfers write ownership.
7. Pick the smallest capable configured model; declare file ownership before parallel writes.
8. Challenge techniques use **fresh context**; agreement is not proof.