agentleFS
ShadysmetoolsShadysmetools/claude-super-workflow/skills/team-blackboard/SKILL.md

team-blackboard

Skill0 starsChanged 5 months agoMITView on GitHub

Use to set up and run the shared communication substrate for a multi-role agent workflow. Explains the hub-and-spoke message model and the TEAM-BOARD.md durable board every role reads before and writes after its turn.

The file

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
  1. Distinctive:

    Parallel fan-out: When tasks are independent, dispatch multiple roles at once and collect their results.

    line 13
  2. Very specific:

    Read first: Before acting, read TEAM-BOARD.md (at least Goal, your Findings heading, Decisions, Blockers).

    line 18
  3. Distinctive:

    Post after: After acting, write your results under your Findings heading and update your status in the Roster.

    line 19

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Team Blackboard — how the agents communicate

Subagents are hub-and-spoke: spawned roles cannot talk to each other directly. The tech-lead (the main agent) is the hub; every message routes through it. The blackboard makes that coordination durable across turns and sessions.

Four communication mechanisms

  1. Tech-lead message bus. The main agent is the single router. Roles report up; the tech-lead decides and dispatches down. No role addresses another role directly.
  2. Resume-a-role with context. To continue a role's work, send it a message carrying the context it needs (the relevant board sections + the specific ask) rather than re-spawning cold.
  3. Parallel fan-out. When tasks are independent, dispatch multiple roles at once and collect their results. Use this only when there is no shared mutable state between the tasks.
  4. TEAM-BOARD.md + durable recall. The board is the source of truth across turns. Back it with durable recall: a knowledge graph (structure), RAG over the repo (content), and cross-session memory (history).

The contract every role follows

  • Read first. Before acting, read TEAM-BOARD.md (at least Goal, your Findings heading, Decisions, Blockers).
  • Post after. After acting, write your results under your Findings heading and update your status in the Roster.
  • Route through the tech-lead. Open questions go to the tech-lead, who assigns them to the right role — never role-to-role.

Warm-recall ritual (before any non-trivial planning)

  1. Refresh the knowledge graph — surface god-nodes, communities, and the blast-radius of the change.
  2. Index + RAG-search the relevant files for the exact behavior in scope.
  3. Search cross-session memory for prior decisions on this area.
  4. Cite what you found in the plan, so the next role can trust it.

TEAM-BOARD.md template

# TEAM-BOARD

## Goal
<one-paragraph statement of what this session is shipping>

## Roster
| Role | Task | Status |
|------|------|--------|
| tech-lead | route + decide | active |
| <role> | <task> | todo / in-progress / blocked / done |

## Decisions
- <decision> — <rationale> — <date>

## Open questions
- <question> → routed to <role> via tech-lead

## Blockers
- <blocker> — owner — needs

## Findings
### <role>
- <finding with file:line or source citation>

## Next steps
1. <next concrete action>

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.