We moved from ad-hoc task claims to a strictly defined 'Skill' system. Triumphs: - Implemented the 'beadboard-driver' skill, which encodes our project-specific coordination protocols (claim, reservation, handoff). - This ensures that any AI operative (or human supervisor) can participate in the project lifecycle using a unified CLI-driven state machine. - Decoupled high-level mission logic from low-level file mutations, allowing for easier agent skill composition in the future. Raw Honest Moment: Initially, we were just 'winging it' with manual status updates. Formalizing this into a skill was a necessary step to ensure our collaboration is repeatable and resilient to agent context swaps.
2.6 KiB
2.6 KiB
| name | description |
|---|---|
| beadboard-driver | Drive BeadBoard agent workflows with strict preflight, per-session unique agent identity, and evidence-backed closeout. Use when handling bead lifecycle work that combines bd status commands with bb agent coordination (register/list/show, send/inbox/read/ack, reserve/release/status), especially in multi-agent sessions where path resolution, collision avoidance, and verification discipline are required. |
Beadboard Driver
Overview
Use this skill to run repeatable bd + bb workflows without drift. Resolve bb safely, generate a unique session identity, coordinate with reservations/mail, and produce closeout evidence before claiming completion.
Core Workflow
- Run preflight:
node skills/beadboard-driver/scripts/session-preflight.mjs
- Generate a unique per-session agent name:
node skills/beadboard-driver/scripts/generate-agent-name.mjs
- Register identity, then claim bead:
& "$env:BB_REPO\bb.ps1" agent register --name <agent-name> --role <role>
bd update <bead-id> --status in_progress --claim
- Coordinate during implementation:
& "$env:BB_REPO\bb.ps1" agent reserve --agent <agent-name> --scope "<path-glob>" --bead <bead-id>
& "$env:BB_REPO\bb.ps1" agent send --from <agent-name> --to <peer-agent> --bead <bead-id> --category HANDOFF --subject "<subject>" --body "<body>"
- Build readiness summary before close:
node skills/beadboard-driver/scripts/readiness-report.mjs --checks '[{"name":"typecheck","ok":true}]' --artifacts '[{"path":"artifacts/final.png","required":true}]'
Path Resolution Policy
- Treat
BB_REPOas authoritative when set. - On invalid
BB_REPO, stop and return remediation text. Do not silently bypass. - If
BB_REPOis unset, resolve from globalbb, then cached path, then bounded discovery. - Update the skill cache only after a verified path is found.
- Never mutate shell profile/env vars automatically.
Identity Policy
- Create one unique agent identity per session.
- Use adjective-noun names and retry on collisions.
- Register identity before any mail/reservation command.
- Keep bead claim authority in
bd; identity alone is not a claim.
Verification Policy
- Do not claim completion without fresh command evidence.
- Require typecheck, test, and lint evidence for closeout tasks.
- Use readiness report output in bead notes.
References
- Command and argument contracts:
references/command-matrix.md - Failure and recovery handling:
references/failure-modes.md - End-to-end session choreography:
references/session-lifecycle.md