AGENT SKILL

review-triage

Use when the user asks to review the work queue, triage pending items, clear the backlog, or find out what needs attention in the vault. Surfaces prioritized actionable items and batch-processes approved updates.

License 许可
MIT
最近更新
2026-03-22
Stars
★ 0
Forks
0

01 前言 Front Matter

name
review-triage
version
1.0.0
description
Use when the user asks to review the work queue, triage pending items, clear the backlog, or find out what needs attention in the vault. Surfaces prioritized actionable items and batch-processes approved updates.
disable-model-invocation
true

02 资源文件 2

03 SKILL.md 正文

Work Queue Review and Triage

Review the vault's work queue, evaluate each item, and batch-process approved actions. This skill uses the loop pattern: enumerate all items, inspect each, propose actions, then execute the approved set after user confirmation.

Iron Laws

Never write without confirmation. Present the full proposed action set BEFORE executing any writes. The user approves or prunes the batch before a single write fires.

Check result.success after every write. If update_content or close_content fails, stop the batch and report what succeeded and what failed. Do not silently skip failures.

Workflow

  1. Check vault integrity — check_integrity(): run the 4-category integrity scanner (orphan edges, broken links, missing files, schema issues). Surface any critical issues before reviewing the work queue. If integrity issues are found, report them first — they may explain items in the queue.

  2. Load queue — work_queue(): load the prioritized work queue. All actionable items are returned with composite scores. If the queue is empty and no integrity issues, report "Work queue is empty. Vault is clear." and stop.

  3. Inspect top items — For each of the top items (up to 10): get_document(content_id="<id>") — fetch full content to evaluate. Parallel fetches are acceptable.

  4. Evaluate each item — Classify and propose an action:

    • Stale seed: maturity=seed, age > 7 days, no body updates → suggest: promote to budding or archive
    • Actionable task: has a clear next step, not yet done → suggest: keep as-is or mark complete
    • Orphan note: 0 outgoing links, status=draft → suggest: add links via reweave, or archive
    • Draft needing attention: draft status, has content but no connections → suggest: add tags/links to promote
  5. Present summary table — Show proposed actions before any writes:

    | # | ID | Title | Type | Suggestion | Priority |
    

    Ask: "Process all / only high-priority (score > X) / pick specific items?"

  6. Batch execute — For each approved item, execute the proposed action:

    • Promote maturity: update_content(content_id="<id>", changes={"maturity": "budding"})
    • Mark complete: update_content(content_id="<id>", changes={"status": "complete"})
    • Archive: close_content(content_id="<id>")
    • Reweave orphan: reweave(content_id="<id>") Check result.success after each call. Stop the batch on failure, report progress.
  7. Report results — Summarize: integrity issues found, items reviewed, items updated, items archived, items remaining in queue.

Batch Confirmation Pattern

The confirmation model prevents accidental bulk writes:

  • Present first, execute second — show the full proposed action set, then wait for approval
  • Scope selection — user can select "all", "high-priority only" (score above a threshold), or name specific IDs
  • Archive warning — archives are highlighted separately; close_content is a soft-delete that hides the item from default queries
  • After batch — report successes and failures; surface any items that need manual review

See references/triage-workflow.md for item evaluation criteria and priority scoring details.

04 同仓库更多 Skills

ThatDevStudio
ThatDevStudio/ztlctl

vault-methodology

Use when working in a ztlctl vault — creating notes, organizing knowledge, tagging, linking, or making decisions about content types and structure. Activates for any knowledge management task.

★ 0
ThatDevStudio
ThatDevStudio/ztlctl

synthesize

Use when the user wants to synthesize scattered notes into a unified artifact, consolidate knowledge on a topic, connect existing notes, find structural gaps, or create a synthesis note. Activates on "synthesize", "consolidate", "connect my notes on X", "summarize what I know about X", "find gaps in X", or "create synthesis".

资料检索与研究
★ 0
ThatDevStudio
ThatDevStudio/ztlctl

session

Use when the user wants to start, begin, or open a work session, OR close, end, or finish an active session. Manages the full session lifecycle: pre-flight checks, polaris alignment, session start, and session close with enrichment pipeline.

★ 0
ThatDevStudio
ThatDevStudio/ztlctl

session-workflow

Use when starting, running, or closing research sessions. Guides structured knowledge capture workflows: start → capture → synthesize → close.

资料检索与研究Skill与智能体开发
★ 0
ThatDevStudio
ThatDevStudio/ztlctl

review-contradictions

Use when the user wants to review conflicting notes, find inconsistencies in the vault, check for contradictions between knowledge claims, or inspect semantically-close notes that may conflict. Runs the per-pair evaluation loop to surface genuine contradictions and dismiss false positives.

★ 0
ThatDevStudio
ThatDevStudio/ztlctl

orient

Use when entering a vault cold, starting any work, or needing to understand what is in the vault. Orients to vault identity, strategic priorities, and assembles topic-focused context. Activates on "orient", "load context", "what is in the vault", or "vault status".

★ 0