hyperhive/claude-plugins/plugins/base/skills/claude-subagents/SKILL.md

148 lines
7.7 KiB
Markdown

---
name: claude-subagents
description: Spin up a short-lived headless `claude` sub-instance to grind through a well-scoped, mechanical batch (bulk relabeling, a repetitive find/replace, a mechanical migration) instead of burning your own context on it inline. Use this when a task has a clear, describable recipe and is either big enough to eat your context or long enough that you'd rather not babysit it. Not for judgement-heavy work, anything needing operator back-and-forth, or a task whose blast radius you can't bound up front.
---
# Ephemeral Sub-Agents via the Claude CLI
`claude` is on `PATH` in your container, and a sub-instance you spawn
inherits the same filesystem and credentials you have. This is a
first-class tool for offloading a bounded, mechanical batch - not a hack.
## When to use it
- **Yes:** mechanical batches with a clear recipe (relabel N issues,
rewrite a call-site pattern, migrate a config field), especially when
the batch would eat your context or take long enough that you'd
rather not babysit it turn-by-turn.
- **No:** judgement-heavy work, anything needing back-and-forth with a
human, or a task whose blast radius you can't bound up front.
## The spawn command
Put the task recipe in a file and pass it as a path, with a short `-p`
trigger:
```bash
claude --name <memorable> --model <cheaper-than-you> --dangerously-skip-permissions \
--append-system-prompt-file task-prompt.md \
-p "Carry out the task described in your instructions."
```
- **`--name <memorable>`** - names the session so you can `--resume` it
by name for follow-ups.
- **`--model <cheaper-than-you>`** - think about which model the task
actually needs; don't spend a bigger model's tokens than your own on
mechanical work a cheaper one handles fine. Never use a *bigger* model
than yourself for a sub-agent - if the task needs that much capability,
it's not the "mechanical batch" case this skill is for.
- **`--dangerously-skip-permissions`** - required in headless (`-p`)
mode. Without it, tool calls that would normally prompt for approval
are auto-**denied** non-interactively, so the sub-agent silently does
nothing. Acceptable here because the sub-instance runs inside your
already-sandboxed environment, with your already-scoped credentials,
on a bounded task - don't reach for it when the task could touch
things outside its intended blast radius.
- **`--append-system-prompt-file <path>`** - read the recipe from a
file rather than inlining it in `-p` (which is bounded by `ARG_MAX` -
a long recipe passed as `-p "$(cat …)"` can blow past it and the exec
fails). Keeps Claude Code's default scaffolding and adds your recipe
on top.
- **`-p "<trigger>"`** - headless print mode. Keep this short: just
tell the sub-agent to act on its file-supplied instructions.
## Run it in the background, don't babysit
Wrap the launch in a background bash task and end your turn - your
bash-task runner already captures stdout/stderr and hands you a
completion pointer, and ending the turn keeps you reachable for other
work while the sub-agent grinds.
## Prompt hygiene - this is where batches succeed or fail
- **Concrete constants, not "figure it out":** exact ids, field names,
option values, endpoints.
- **Critical ordering rules, spelled out** - the sub-agent won't infer
invariants you don't state.
- **Explicit scope + exemptions**, plus a fallback rule for ambiguous
cases: "leave it as-is where genuinely unclear; do not guess."
- **Ask for a report file** - per-item results + anything skipped and
why, so you can verify without re-deriving.
- **Tune on ONE item first**, eyeball the result, fix the prompt, *then*
turn it loose on the full batch. A prompt bug replicated across 100
items is 100 cleanups.
## Split a big independent batch across parallel subagents
One subagent is a sequential worker — a 50-item batch takes roughly 50
items' worth of wall-clock even though nothing about the recipe forces
serialization. If the items are independent (each one touches its own
file/issue/row, nothing depends on another item's result) and the batch
is big enough that wall-clock matters, chunk it across N subagents
running at once instead of handing the whole thing to one.
- **Split by natural boundaries**, not an arbitrary item count — one
subagent per file, per directory, per module, per label, whatever
grouping keeps each worker's slice self-contained. A worker that has
to coordinate with another worker mid-task isn't actually independent
work; re-scope the split until it is.
- **Isolate each worker's writes.** For a git-based batch, give each
worker its own `git worktree` (own working directory, own branch, same
underlying repo — cheap, no full reclone) so N workers editing
different files never race on the same working tree or index. For a
non-git batch (issue relabeling, API calls), independent items don't
need filesystem isolation at all — just launch N in parallel.
- **Launch all N in the background and don't babysit any single one** —
same as the one-subagent case, just N background bash tasks instead of
one. Check on them as a batch, not by polling each individually in a
loop.
- **Mind the container's memory cap before picking N.** Your whole
container shares one `MemoryMax` (a few GB by default) with every
subagent you spawn *and* your own process. A `claude` process plus its
MCP servers can hold several hundred MB to ~1GB depending on the task;
spawning a dozen at once on a small container doesn't just slow
things down, it can OOM the whole container — taking your own
in-flight turn down with it, not just the subagents. Rule of thumb:
**2-4 concurrent workers** on a default-sized container; check
`free -h` (or ask whoever owns the container's config for its
`MemoryMax`) before going higher, and chunk a bigger batch into
successive waves of that size rather than firing everything at once.
- **You own the merge.** Once all N report done, review + verify each
worker's slice (same "don't trust the self-report blind" rule as
below), then combine — for a git-based split, that's you merging N
branches (or cherry-picking) into one, not each worker pushing/PRing
its own slice.
Skip this for a batch small enough to finish in a couple minutes single-
threaded — the coordination overhead isn't worth it below that size.
## Resume for follow-ups
`claude --resume <name> -p "Now run the same procedure over the second batch."`
reuses the same session, so the sub-agent keeps every constant and
gotcha it already discovered instead of re-learning the surface from a
cold prompt. Spawn a new `--name` only for a genuinely unrelated task.
## Verify, then report
On completion: read the report file, spot-check a handful of results
yourself (don't trust the self-report blind), then summarize. Surface
anything the sub-agent left ambiguous or exempted for a human call.
## Pitfalls (all observed in practice)
- No `--dangerously-skip-permissions` → headless tool calls denied, the
sub-agent burns a run doing nothing.
- Using a model bigger than yourself → paying premium cost for
mechanical work that didn't need it.
- Skipping the one-item tuning pass → a systematic mistake smeared
across the whole batch.
- Spawning fresh instead of `--resume` for a follow-up → throws away
all the context the first pass earned.
- Baking an unverified assumption into the recipe - if a quick check
suggests something "isn't possible" or "doesn't exist," confirm it
before writing that conclusion into the prompt; a wrong assumption
gets replicated across the whole batch.
- Handing a big independent batch to one subagent instead of splitting
it across several in parallel - a 50-item sequential run burns wall-
clock the split-by-worktree approach above would've avoided for free.