Files
skills/skills/kvb-org-perform/SKILL.md
T
Kevin Veen-BirkenbachandClaude Opus 5.5 8bc7209ed1 feat(orchestrator): review staged work as match cards (requirement 012, phases A and B)
Requirement 012 plus its first two phases: the kvb-match skill (match.py add, list, show) feeds the existing review queue, /api/match serves the cards and local-only stale and snapshot posts, and payload files are read only as regular files under /tmp within the sandbox view and 64 KiB. The card at #tasks/match shows face, name, age in days, repository, branch, worktree and files that open while scrolling; right-click or long press marks a line, stale cards cannot be accepted. Accept, reject and the passkey release follow in phase C after the screenshot feedback. Tests run once on unify/skills-20261114 after the merge (operator order 2026-11-14).

🎯 Fix: 75 %
💥 Break: 8 %
💸 Tokens: 100 %
⏱️ Velocity: 100 %
🛡️ Security: 95 %
🏆 Score: 66
🚦 Gate: ⛔ fix 75 % < 90, break 8 % > 5; security 95 % < 100: ask the operator before the commit

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-14 21:36:40 +00:00

3.6 KiB

name, description
name description
kvb-org-perform Pick `model` and `effort` per subagent from a fixed table, cheaper only at 90% confidence of equal results. Apply AUTOMATICALLY before every Agent call.

A subagent runs on the main model and effort unless the spawn names others. Set model and effort only for a task that clearly matches a row below, at least 90 % sure the cheaper setting gives the default's result: no lost findings, no rework. Cost counts per finished task, retries included. Otherwise set neither.

Table

Task class Subagent's task model effort Evidence
lookup Read-only: find files, symbols or call sites, grep, list, read and quote, pull a log line or field, summarise a given text without judging it haiku low Haiku 5.5 costs 1/40 of Opus 5.5 per token and is the speed tier; Anthropic names Haiku at low effort for simple, speed-critical and reading-heavy worker tasks; Explore and the cavecrew investigator already run on it
mechanical-edit An edit the brief fully specifies (exact old and new text, a rename, deleting listed lines) in one or two files haiku low nothing is left to decide, and the parent checks the diff
judgement Everything else: root cause, skeptic or dialectic antithesis, review, security or legal checks, design, new or multi-file code, writing tests, performance work omit omit no evidence at 90 % that a cheaper setting gives the same result

Prices per million tokens in and out (claude-api reference, cached 2026-10-06): Haiku 5.5 $0.10 / $0.50, Sonnet 5.5 $2 / $10, Opus 5.5 $4 / $20, Fable 5.1 $10 / $50.

Spawn

  1. Pick the class; a task that mixes classes, or whose class you hesitate over, is judgement.
  2. Start the brief with task-class: <class>; the spawn hook logs it with model, effort and the Agent description.
  3. Set model and effort from the row; for judgement set neither.
  4. Spawn as many agents as the stakes need: one, unless independent parts gain from running in parallel. The brief names the exact files, lines and claims to check and what is ruled out, and caps the answer (for example "at most 15 lines, file:line per finding").
  5. Enter the spawn in the kvb-org-rasci register: the subagent as R and as --worker "subagent <type> (<model>, <effort>)" of the task it serves.
  • An agent type whose definition sets its own model (Explore, Plan, the cavecrew agents) keeps it; never pass model or effort for it.
  • Never pass fable or opus: neither is cheaper than the default.
  • A Haiku agent that reports the task needs judgement, or returns an incomplete result, is not retried on Haiku: the default continues it.
  • Never change the default itself (settings, CLAUDE_CODE_SUBAGENT_MODEL, agent definitions); this skill only chooses per call.

Feedback

Once you have judged the result, rate it under the same Agent description:

${CLAUDE_SKILL_DIR}/scripts/log.py feedback --description "<description>" --completed yes|no --tests pass|fail|none --redo yes|no

--redo yes when the default had to redo or finish the work. When the operator later asks a follow-up about that result or corrects it:

${CLAUDE_SKILL_DIR}/scripts/log.py followup --description "<description>" --text "<question or correction>"

All sessions share ~/.local/state/agents-council/subagent-model.jsonl; a sandbox that cannot write it gets the line printed as pending:, and the PostToolUse hook stores it.

Read evaluation.md next to this file before reading log.py report or changing a row of the table.