snappy-launch

Orchestrate parallel subagent launches with coordinated work tracking. This skill pairs with snappy-subagents (which governs an individual subagent's behavior) to handle the higher-level orchestrator's responsibilities: deciding how many parallel agents, what each does, collecting their writebacks, and assessing whether the parallel work made aggregate progress. Failure mode prevented: orchestrators that launch agents but don't verify they wrote back, or that launch N agents without confirming the work was actually independent.

Steps

1. Define work units

• Identify independent tasks that can run in parallel (fan-out shape).
• Each unit is one Task or headless claude -p subagent.
• Skip the launch if tasks touch the same files (can't parallelize safely).

2. Compose subagent briefs

• For each unit, write a brief that names the setpoint (state/skills/<slug>/AGENTS.md).
• Include explicit FORBIDDEN paths list (files this subagent must NOT touch).
• Mandate writeback before exit: if the loader has a gap, edit this AGENTS.md inline. git log on the file is the audit trail.

3. Launch and collect

• Fan out the subagents (Task tool for attended, headless claude -p for Stop-hook drains).
• Wait for all subagents to return and emit writebacks.
• Read the feedback log; verify each subagent's writeback is present.

4. Assess progress

• For each subagent's writeback, check if the claimed [FIXED] edit is on disk.
• Tally: how many succeeded, how many need re-dispatch, how many silently failed.
• Log an overall verdict (1.0 = all clean | 0.5 = some re-dispatch | 0.0 = any silent returns).

5. Log + eval

import { score } from "../../lib/eval";
score("snappy-launch", run_id, {
  score: <1.0 if all subagents wrote back AND every [FIXED] claim on disk
         | 0.5 if writebacks present but ≥1 re-dispatch needed
         | 0.0 if any silent return OR clobber OR lied [FIXED]>,
  primary_issue: <null | "silent-return" | "clobber" | "lied-fix">,
});

Eval

Actor: the thing that produces the output (a dispatch model, a CLI, or the state/lib/snappy-launch.ts library if one exists). Auditor: the thing that judges (must be different - see CONSTITUTION invariant #3). Name both explicitly.

Score convention:

If you cannot name a deterministic auditor, switch the frontmatter to eval: manual and log to state/log/pending-eval.ndjson - but fight to avoid manual. Manual is the escape hatch that leaks the thesis.

Gotchas

• List concrete failure modes you hit while building. Examples: "API

returns 200 with empty body"; "round-tripped text has a , artifact where em-dashes used to be"; "cache file is stale after 12h."

• If a gotcha matches an entry in user memory (see ~/.claude/CLAUDE.md),

cite the memory name so future agents can follow the trail.

snappy-launch - loader

Parallel subagent orchestration. Full reference: state/skills/snappy-launch/SKILL.md.

Critical Rules

1. Define work units. Identify truly independent tasks (fan-out). Each = one Task (attended) OR one claude -p subagent (Stop-hook). Skip if tasks share files - parallelism requires disjoint file sets.
2. Compose briefs with FORBIDDEN paths. Name setpoint (state/skills/<slug>/AGENTS.md). List files agent must NOT touch. Mandate writeback: echo "[$(date -u +%FT%TZ)] <slug>: <gap or fix> [FIXED|LOGGED]" >> state/log/loader-feedback.log.
3. Launch, wait, collect. Fan out (Task for attended | claude -p for drains). Wait for ALL to exit. Read feedback log, verify each writeback present.
4. Assess progress. For each writeback, verify [FIXED] claim on disk. Tally: success | re-dispatch | silent-fail (no writeback). Silent failures = score 0.0.
5. Score eval row. score() from state/lib/eval. 1.0 = all wrote back AND all [FIXED] on disk | 0.5 = writebacks + ≥1 re-dispatch | 0.0 = silent return OR clobber OR lied. Log primary_issue (null | "silent-return" | "clobber" | "lied-fix").
6. Actor ≠ Auditor. You launch. Distinct audit step grades (CONSTITUTION #3).
7. Default scope-only (apply:false). Preview work. Explicit confirmation before side effects.

Commands

| define | identify truly independent Task / claude -p per work unit | | compose | name state/skills/<slug>/AGENTS.md, list FORBIDDEN paths, mandate writeback line | | fan out | Task(name="<unit>", prompt="<brief>") attended; claude -p for drains | | collect | grep "<slug>" state/log/loader-feedback.log to verify all writebacks | | verify | ls -l <file> per [FIXED] claim to confirm edits on disk | | score | score("snappy-launch", run_id, { score, primary_issue }) → state/log/evals.ndjson |

Self-Test

• [ ] Name independent work units (disjoint file paths)?
• [ ] Include FORBIDDEN paths in every brief?
• [ ] Wait for ALL subagents before scoring?
• [ ] Verify [FIXED] claims by reading files?
• [ ] Score if one of three never writes back?
• [ ] Distinguish 0.5 vs 0.0?
• [ ] Actor AND auditor, or separate grader?

Found a gap? Edit this file. <!-- footer-injection-point -->