Skip to content

Add roster-tuner skill: agency roster performance tuning#171

Open
tttt28444 wants to merge 20 commits into
runxhq:mainfrom
tttt28444:skill/roster-tuner
Open

Add roster-tuner skill: agency roster performance tuning#171
tttt28444 wants to merge 20 commits into
runxhq:mainfrom
tttt28444:skill/roster-tuner

Conversation

@tttt28444

Copy link
Copy Markdown

Summary

Adds the skill to the runx catalog.

This skill reads a sealed agency case event stream through the hosted data-store, folds per-member metrics (turn count, refusal tally, completion time), ranks members against operator-supplied norms, and emits a typed roster tuning decision naming the bounded member change as plain data.

What it does

  • Reads case projection via ()
  • Folds per-member metrics: turn count, refusal rate, avg completion time
  • Grades each member against operator norms (refusal_threshold, completion_time_threshold)
  • Enforces guard rails: min_roster_size, sole_skill protection, schema version match
  • Appends judgment as ungated CAS write to the case event stream
  • Emits decision packet; never executes a roster mutation

Harness cases

  • : member at refusal_rate 0.75 above 0.6 threshold and completion_time 3x norm, yields underperformer verdict with skill-matched replacement
  • : caller.answers omitted, grading agent-task blocks, run stops with needs_agent before any decision is appended

Files

  • skills/roster-tuner/X.yaml - skill definition with typed inputs/outputs and graph runner
  • skills/roster-tuner/SKILL.md - human/agent-readable skill documentation
  • skills/roster-tuner/fixtures/ - harness test case fixtures
  • skills/roster-tuner/graph/ - sub-graph implementations (fold, grade, decide)

Usage

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants