用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt;运行前请先检查来源。
npx skills add https://github.com/majiayu000/claude-skill-registry --skill knowledge-activation命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地?可下载 SkillsMP 当前能够提供的文件。
LLM token logprobs and calibration. Per-decision confidence, ECE, Brier, reliability diagrams, low-confidence triage.
Analyze LLM token logprobs and calibration. Use for per-decision confidence, ECE, Brier scores, reliability diagrams, and low-confidence triage.
回顾最近 N 天的 Claude Code 使用记录——扫描原始会话数据,按主题分组汇总"我都做了什么",并从个人操作系统视角输出模式、风险与增删建议。当用户说 /recap、"看看我这几天做了什么"、"回顾一下我最近的会话"、"这两天我用 claude 干了啥"、"活动回顾" 时使用。
基于 SOC 职业分类
正在显示 SKILL.md
| name | knowledge-activation |
| description | Activate mature .agents knowledge. |
| practices | ["wiki-knowledge-surface","pragmatic-programmer"] |
| hexagonal_role | supporting |
| consumes | [] |
| produces | [] |
| context_rel | [] |
| skill_api_version | 1 |
| user-invocable | true |
| context | {"window":"fork","intent":{"mode":"task"},"sections":{"exclude":["TASK"]},"intel_scope":"topic"} |
| metadata | {"tier":"knowledge","dependencies":["compile","harvest","flywheel"]} |
| output_contract | .agents/beliefs.md, .agents/playbooks/*.md, .agents/briefings/*.md |
Turn a mature .agents corpus into operator-ready knowledge surfaces.
Use this skill when the problem is no longer "capture more knowledge," but:
$compile remains the hygiene loop. knowledge-activation owns corpus operationalization.
Knowledge activation is the fourth step in the global-corpus workflow:
$harvest — gather artifacts from many rigs into ~/.agents/learnings/$compile — synthesize raw artifacts into .agents/compiled/$dream overnight — bounded compounding loop$knowledge-activation — lift compiled knowledge into playbooks,
beliefs, and runtime briefingsSee docs/skills-decision-tree.md for the full "which skill next?" decision table covering harvest, compile, dream, knowledge-activation, and quickstart.
This skill assumes the current workspace already has:
.agents/ directory.agents/scripts/ when ao knowledge activate needs to rebuild source manifests, topics, promoted packets, and chunk bundles from custom workspace logic.agents/harvest/latest.json, which ao knowledge activate can use as a native fallback to turn the latest harvest catalog into a harvested-praxis topic packet, promoted packet, and chunk bundleRead references/script-contracts.md for the required builder inventory and command ownership.
The stable product surface is the ao knowledge command family:
ao knowledge activate --goal "turn agents into usable information"
ao knowledge beliefs
ao knowledge playbooks
ao knowledge brief --goal "fix auth startup"
ao knowledge gaps
The skill owns routing, sequencing, interpretation, and next-step recommendations. ao owns the belief/playbook/brief/gap product surfaces directly.
ao lookup and ao codex start consume these outputs as operator context. Matched knowledge briefings are the preferred dynamic startup surface, while selected beliefs and healthy playbooks provide bounded supporting guidance.
Verify that .agents/ exists. When you plan to run ao knowledge activate, verify that at least one evidence substrate is present:
source_manifest_build.py, topic_packet_build.py, corpus_packet_promote.py, knowledge_chunk_build.py.agents/harvest/latest.jsonao knowledge beliefs, ao knowledge playbooks, ao knowledge brief, ao knowledge gapsRun the packet layers in order:
Read references/dag.md for the full DAG and its trust gates.
Refresh the promoted operator layers:
ao knowledge beliefs
ao knowledge playbooks
These should materialize the consumer surfaces under .agents/knowledge/ and .agents/playbooks/.
When there is an active objective, compile a bounded startup aid:
ao knowledge brief --goal "your goal here"
The briefing should stay small, cite its source surfaces, and include warnings when a selected topic is thin.
Run:
ao knowledge gaps
This reports thin topics, missing promotions, weak claims needing review, and the next recommended mining work.
If you want the complete pass in one step, run:
ao knowledge activate --goal "your goal here"
That command sequences evidence consolidation, belief/playbook refresh, optional briefing compilation, and a gap summary.
Read references/output-surfaces.md for the canonical output surfaces and trust boundaries.
The consumer-facing outputs are:
.agents/knowledge/book-of-beliefs.md.agents/playbooks/index.md.agents/playbooks/<topic>.md.agents/briefings/YYYY-MM-DD-<goal>.md.agents/retro/The substrate surfaces remain:
.agents/packets/.agents/topics/.agents/packets/chunks/catalog.jsonlActivate the full outer loop for an active goal
/knowledge-activation
ao knowledge activate --goal "productize knowledge activation"
Refresh only the belief and playbook promotion layers
ao knowledge beliefs
ao knowledge playbooks
Check whether the corpus is safe to promote
ao knowledge gaps