AB frugal-subagents
On Claude Fable 5.1 or Opus, subagents inherit the session's expensive model and can spawn more of themselves, so one research fan-out can burn the usage limit in minutes; this plugin's PreToolUse hook runs every subagent on a cheap model unless one is named explicitly, blocks nested spawns and caps spawns per session, and this skill says how to delegate well under those rules. Use whenever the session is about to delegate — spawning agents, running web scans or research passes in parallel, "fan out", "run a fleet of agents", "search across many sites", comparing listings/prices/flights/apartments/suppliers across the web — and whenever the user says subagents burn through the limit, agents spawned more agents, asks to "limit subagents", "run agents on a cheaper model", "set up frugal subagents", "which model did the helpers use", or when Claude Code reports that the frugal-subagents hook failed to run (Node.js missing).
On Claude Fable 5.1 or Opus, subagents inherit the session's expensive model and can spawn more of themselves, so one research fan-out can burn the usage…
As a process B 72/100 · Nearly there — weak spots: result and completion, inputs and preconditions
How to improve
- Your own cases (evals/evals.json, 4–6 real requests with expected answers): the full check would then run those instead of a model-drafted suite.
- A spec.yaml with trigger phrases and assertions — a behaviour contract for CI; `skilltest init` writes a template.
Guard findings · 0
✓ No critical or high findings
Files scanned: 7. Evidence is masked. Grey chips explain why severity was lowered.
Against the Agent Skills spec
✓ No remarks against the Agent Skills spec
Process rating: all ten parameters 72/100
- 30Inputs and preconditions. Does not say what the process needs to start
- 40Result and completion. Does not say what the result is
- 55Failures and branches. 1 branches
- 60Tools and files. Uses tools (bash, web) that frontmatter does not declare
- 100Steps. 8 steps
- 100When it triggers. States when to use and when not to
- 100Consistency. Name and required fields are in place
- 100Execution cost. Instruction body is 1905 tokens
- 100Running it twice. No mutating operations
- 100Progress reporting. Reports progress
Everything here is measured from the skill text rather than judged by a model, so the numbers are checkable. A parameter weighs more when it is a more common reason for the process to stall.
Quality signals
- +4Description does not say when NOT to use the skill (false activations)
- +3Description length 934: 120–800 characters recommended
- +3Output format is not stated: the model decides each time
- +1No license
- +2Single-language instructions
- +5Description quotes 7 example trigger phrases
- +4Structure: 8 headings
- +3Step-by-step instructions: 8 items
- +4Has examples (1 code blocks)
Quality base 70; lint remarks subtract, signals add up to 100. Result: 86.