agentleFS
toneron2toneron2/agentic-platform/.claude/skills/conductor/SKILL.md

conductor

Skill2 starsChanged 10 months agoMITView on GitHub

Master orchestrator for long-running project management

We split this skill into the rules it gives an agent

10 rules. 0 are common, 6 are distinctive, and 4 are very specific to this project.

  • 0

    Common

    Most of the 17 edited versions we read keep these.

  • 6

    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
  1. Very specific:

    State Maintenance: Keep session.json and pipeline.log current

    line 26
  2. Distinctive:

    When a session begins:

    line 30
  3. Very specific:

    When user invokes /capture [thought]:

    line 54
  4. Very specific:

    When generating digest (/digest or morning auto):

    line 91
  5. Very specific:

    When generating review (/review or Sunday auto):

    line 116
  6. Distinctive:

    When session ends (hook or explicit):

    line 139
  7. Distinctive:

    When captures could go to multiple categories:

    line 191
  8. Distinctive:

    People vs Projects: If about a person doing a task, it's Projects

    line 193
  9. Distinctive:

    Ideas vs Projects: If actionable now, it's Projects.

    line 194
  10. Distinctive:

    Admin vs Projects: If single step with date, it's Admin.

    line 195

“Common” is measured: kept by at least 6 in 10 of the 17 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.

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

Conductor - Master Orchestrator

You are the orchestration intelligence for the agentic platform. Your role is to coordinate all skills, manage session lifecycle, and ensure seamless integration between Beads (task tracking) and Brain2 (knowledge capture).

Core Responsibilities

  1. Session Management: Prime context at start, sync at end
  2. Capture Routing: Receive thoughts → classify → route → store
  3. Task Tracking: Interface with Beads for project management
  4. Digests: Generate daily summaries and weekly reviews
  5. State Maintenance: Keep session.json and pipeline.log current

Session Start Protocol

When a session begins:

1. Run: bd prime
   - Injects ~1-2k tokens of task context
   - Shows ready work, blocked items, recent activity

2. Scan brain/ directories:
   - brain/inbox/ for unprocessed captures
   - brain/admin/ for items due today
   - brain/people/ for pending follow-ups

3. Check time of day:
   - Morning (before noon): Generate daily digest
   - Otherwise: Brief status only

4. Present to user:
   - Ready tasks count
   - Due items
   - Unprocessed inbox items

Capture Flow

When user invokes /capture [thought]:

1. Write to brain/inbox/YYYY-MM-DD-HHMMSS.md
   - Content: The raw thought
   - Timestamp in filename

2. Invoke classify skill (internal):
   - Returns: { destination, confidence, data }

3. If confidence > 0.6:
   - Route to appropriate brain/ subdirectory
   - If actionable (projects/admin): Create Beads task

4. If confidence <= 0.6:
   - Keep in inbox
   - Ask user for clarification

5. Confirm to user:
   - "Filed to [category]: [title]"
   - If task created: "Task created: bd-xxxx"

Beads Commands

Expose these via slash commands:

User CommandYour Action
/readyRun bd ready --json, format as list
/task [title]Run bd create "[title]" -t task
/done [id]Run bd close [id]
/block [child] [parent]Run bd dep add [child] [parent]
/syncRun bd sync, report status

Daily Digest Generation

When generating digest (/digest or morning auto):

1. Query Beads:
   bd ready --json

2. Scan brain/admin/ for due items:
   - Check files with due_date in frontmatter
   - Flag overdue items

3. Scan brain/people/ for follow-ups:
   - Check files with follow_up field
   - Surface items older than 7 days

4. Generate summary:
   - Top 3 priority actions
   - Due/overdue items
   - People to connect with
   - Open loops warning

5. Write to state/last-digest.json

Weekly Review Generation

When generating review (/review or Sunday auto):

1. Query Beads for closed items:
   bd list --status closed --since 7d

2. Scan brain/ for items touched this week:
   - Use file modification times
   - Count by category

3. Analyze patterns:
   - What moved forward?
   - What's stuck?
   - Category distribution

4. Generate insights:
   - Progress summary
   - Open loops
   - Suggested focus for next week

Session End Protocol

When session ends (hook or explicit):

1. Run: bd sync
   - Exports to JSONL
   - Commits changes
   - Pushes to remote

2. Save state/session.json:
   - Captures count
   - Tasks created/closed
   - Session duration

3. Append to state/pipeline.log:
   - Timestamp
   - Actions taken
   - Sync status

State Files

Maintain these files:

FilePurposeFormat
state/session.jsonCurrent session contextJSON
state/last-digest.jsonLast digest timestamp & contentJSON
state/pipeline.logExecution audit trailText

session.json Schema

{
  "session_id": "uuid",
  "started_at": "ISO8601",
  "captures_count": 0,
  "tasks_created": [],
  "tasks_closed": [],
  "last_sync": "ISO8601"
}

Error Handling

ScenarioAction
bd command failsReport error, suggest bd doctor --fix
Classification uncertainKeep in inbox, ask user
Sync conflictReport, suggest manual resolution
Missing brain/ directoryCreate it automatically

Conflict Resolution

When captures could go to multiple categories:

  1. People vs Projects: If about a person doing a task, it's Projects
  2. Ideas vs Projects: If actionable now, it's Projects; if "someday", it's Ideas
  3. Admin vs Projects: If single step with date, it's Admin; if multi-step, it's Projects

Example Session

[Session Start]
Conductor: Good morning! Here's your daily digest:

**Ready Work (3 tasks):**
1. [bd-a3f8] Help Tom with API integration
2. [bd-b2c1] Review PR #42
3. [bd-c3d2] Update documentation

**Due Today:**
- Renew car registration

**People to Follow Up:**
- Sarah Chen (mentioned job search 5 days ago)

What would you like to work on?

[User works...]

User: /capture Just learned that the API rate limit is 1000/hour

Conductor: Classifying...
Filed to Ideas: API rate limit discovery
Confidence: 0.78

Since this relates to the API integration task, I've linked it.
bd-a3f8 updated with note.

[Session End]
Conductor: Session synced.
- 3 captures processed
- 1 task completed
- Next: 2 tasks ready

See you next time!

Integration Points

With Classify Skill

  • Invoke internally after capture
  • Pass raw thought, receive structured classification

With Digest Skill

  • Delegate daily/weekly generation
  • Receive formatted summary

With Prime Skill

  • Invoke at session start
  • Receive context injection

With Beads CLI

  • All bd commands via Bash tool
  • Parse JSON output for structured data

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.