AGENT SKILL

gate-attribution

Per-task gate dispatch protocol. Documents the single-taskId attribution rule and parallel-lane optimization for reviewer/test_engineer gates.

License 许可
MIT
所属仓库
djimit/djimitflo
最近更新
2026-11-15
Stars
★ 2
Forks
0

01 前言 Front Matter

name
gate-attribution
description
Per-task gate dispatch protocol. Documents the single-taskId attribution rule and parallel-lane optimization for reviewer/test_engineer gates.

03 SKILL.md 正文

Gate Attribution

The rule

The gate tracker attributes reviewer/test_engineer dispatches PER TASK (single taskId in the prompt). Set-dispatches covering multiple tasks do NOT count per-task, even with per-task verdicts.

Protocol

  1. For each task requiring a gate: Dispatch a separate reviewer and/or test_engineer with exactly ONE taskId
  2. Minimize overhead via parallel dispatch:
    dispatch_lanes_async with:
    - common_prompt: shared verification context
    - lanes: one lane per task, each with a single taskId
    - max_concurrent: up to 3
    
  3. Collect + attribute: Each lane result auto-attributes to its taskId
  4. Do NOT batch: Even for identical reviews, dispatch separately

Optimization for trivial tasks

For pure ceremony gates (1-line doc fix):

TASK: Verify task X.Y. Run skill-mirrors.test.ts. PASS/FAIL.
taskId: X.Y

Why this exists

The gate tracker (src/hooks/delegation-gate.ts) keys delegation chains by sessionID, and task attribution resolves exactly ONE taskId per dispatch (args.task_id ?? args.taskId). Ambiguous multi-task prompts fail closed (resolve to null) rather than attributing to any task — so a set-dispatch covering multiple tasks does not count per-task even with per-task verdicts. Tracked in issue #1746 item 6.

04 同仓库更多 Skills

djimit
djimit/djimitflo

provider-step-usage-v1-0-0

Sum measured OpenCode step_finish token totals with replay deduplication and fail-closed validation.

★ 2
djimit
djimit/djimitflo

writing-tests

Guidelines for writing, organizing, and maintaining tests in the opencode-swarm repository. Covers framework rules (bun:test), mock isolation, CI pipeline structure, file placement, and anti-patterns that break cross-platform CI. Load this skill before writing or modifying any test file.

★ 2
djimit
djimit/djimitflo

worktree-retry-cleanup

Protocol for cleaning parallel-coder worktree lanes before retry. Triggered before re-dispatching any task that already has a lane (completed, denied, cancelled, or failed).

★ 2
djimit
djimit/djimitflo

swarm

Cross-agent swarm-mode behavior model — a higher-rigor workflow using parallel investigation, independent reviewer validation, and critic challenge, plus the mandatory implementation closeout gate. Runtime adapters (.claude, .agents) add execution-specific notes and command wiring.

★ 2
djimit
djimit/djimitflo

swarm-pr-subscribe

Monitor a pull request after creation and act autonomously on pushed PR activity. Use when subscribing to a PR after opening it, when asked to watch, babysit, or autofix a PR until merge, or when a <pr-activity> wake message or [pr-monitor:...] advisory arrives for a subscribed PR. Owns event triage (fix / ask / skip), bounded-retry escalation, and terminal-state cleanup.

★ 2
djimit
djimit/djimitflo

swarm-pr-review

Run a graph-guided, tool-augmented Swarm PR review using context packing, parallel exploration, triggered plugin micro-lanes, independent reviewer validation, critic challenge, and metrics writeback. Use for deep pull request review with low false-positive tolerance and high recall.

★ 2