Skills

Skills (recipes) are slash commands that define what to do. Brewtools includes 13 skills focused on text processing, security scanning, SSH management, deployment, plugin management, subagent budgeting, and session orchestration — each works independently with no prior setup required.

Summary table

#	Skill	Command	Model	Purpose
1	agent-deadline-setup	/brewtools:agent-deadline-setup	sonnet	Install a soft wall-clock budget for subagents — warns at 80%, blocks non-finalizing tools past 100%
2	agent-return-setup	/brewtools:agent-return-setup	sonnet	Installs a size budget on every subagent’s final return — blocks once past threshold, orders compress or write-to-report
3	agent-router-setup	/brewtools:agent-router-setup	sonnet	EXPERIMENTAL. Denies a generic subagent spawn when a real project/plugin expert owns the task, naming it
4	context-slim	/brewtools:context-slim	opus	Compresses everything that permanently enters the LLM context — CLAUDE.md, rules, agent descriptions, hooks, memory — via cross-layer dedup and per-file compression
5	deploy	/brewtools:deploy	opus	GitHub Actions deployment: workflows, releases, GHCR, CI/CD with safety gates
6	manager-setup	/brewtools:manager-setup	sonnet	Project delegation wall and prompt contracts; +++ adds anti-drift cron planning only in verified Plan mode
7	plugin-update	/brewtools:plugin-update	sonnet	Checks, installs, updates Claude Code plugins
8	provider-switch	/brewtools:provider-switch	opus	Configure alt API providers: DeepSeek, Z.ai/GLM, Qwen, MiniMax, OpenRouter
9	secrets-scan	/brewtools:secrets-scan	sonnet	Scans git-tracked files for leaked secrets
10	ssh	/brewtools:ssh	opus	SSH server management — connect, configure, deploy, administer Linux servers with safety gates
11	task-board-setup	/brewtools:task-board-setup	opus	Board, domain/task methodology, work graph, and session anti-drift timers; optional specs and CLAUDE.md optimization
12	text-human	/brewtools:text-human	sonnet	Humanizes code, docs, articles, reddit/chat, javadoc — strips AI artifacts, fixes unicode, fits register
13	text-optimize	/brewtools:text-optimize	sonnet	Optimizes text/docs for LLM token efficiency

How these skills delegate

Delegated work follows the same sizing rule: one subagent = one bounded unit — one deliverable, roughly 5 files and 10 steps. Bigger work is split into bounded tasks and independent owners work in parallel.

Every spawn prompt carries six fields — GOAL, ROLE, SCOPE, CONTEXT, CONSUMER, DONE. A bare one-line task is never enough.

Skill	One bounded unit is…
agent-deadline-setup	one mode-and-scope combination — a single brewcode:hook-creator spawn per combination
agent-return-setup	one mode-and-scope combination — a single brewcode:hook-creator spawn per combination
agent-router-setup	one mode — a single brewcode:hook-creator spawn per mode
context-slim	one file — a single text-optimizer spawn per file, all in one message
deploy	one deliverable, one repo, one environment — confirmation gates stay in the main conversation
provider-switch	one provider — five research agents, one message
secrets-scan	one chunk of the file list — a bigger repo gets more chunks, not bigger ones
ssh	one deliverable on one host — DELETE/PRIVILEGE approval happens before any spawn
task-board-setup	one doc group — same rule at both spawn points
text-human	one block of files from the commit or folder
text-optimize	one file — a folder run is split one file per agent

plugin-update runs its update commands in the main conversation. Manager-setup can delegate a task while keeping wall configuration in the main session.

Task-board prepares methodology, base work units, and a complete unique prompt for every accepted task before implementation. The main session manages one timer per active top-level task, hourly unless you override it; child work units share the parent timer. Planning remains read-only, including when standalone +++ adds a future timer step.

Regression suites

v6 adds automated regression suites where several skills had none. Counts are TOTAL: X | PASS: X | FAIL: 0 from each skill’s own tests/run.sh (tests/suite.mjs for manager-setup), re-run before publishing this table.

Skill	Checks
agent-return-setup	803
agent-deadline-setup	169
agent-router-setup	76
deploy	76
ssh	74
secrets-scan	82
text-optimize	51
manager-setup	44