菜单
Token and prompt-cache discipline — lean always-on rules, stable prefixes, skill indirection, minimal Task payloads, bounded reads.
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。
基于 SOC 职业分类
Track agent decisions, metrics, token costs, and generate reports. Use when logging task execution, tracking agent performance, or generating observability reports.
Unified ai-brain skill combining memory operations and KB query operations with promotion flow.
Bootstraps a Jira hierarchy snapshot before any multi-issue create / edit session. Disambiguates the parent input (issue key, spec URL, Confluence page URL, or `greenfield`), walks parents up (depth-cap 4), walks children down via JQL recursion (depth-cap 4, max 50 children per node), and returns a structured tree alongside the issue-type scheme and link-type metadata the caller needs to render a draft. Used primarily by `atlassian-pm`; secondary consumers (after the G2.5 retrofit) are the upstream Atlassian plugin skills `spec-to-backlog` and `capture-tasks-from-meeting-notes`. The skill **does not render**; rendering belongs to the caller (typically `atlassian-pm` rendering a markdown tree to chat).
Ultra-compressed communication mode. Cuts token usage ~75% by speaking like caveman while keeping full technical accuracy. Supports intensity levels: lite, full, ultra. Defaults: main chat = lite, subagents = ultra. Use when user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman. Also auto-triggers when token efficiency is requested.
Generate → execute → fail → analyze → fix → repeat cycle with failure pattern recognition. Use when implementing changes that need verification and automatic recovery.
Heuristic context budget checks using file-size and count proxies — integrates with context-cache-discipline; no tokenizer introspection claims.
| name | context-cache-discipline |
| description | Token and prompt-cache discipline — lean always-on rules, stable prefixes, skill indirection, minimal Task payloads, bounded reads. |
| version | 1 |
Use this skill when orchestrating or editing the pack so context stays small and early prompt prefixes stay stable (better cache hit odds on providers; always fewer input tokens).
| Goal | Practice |
|---|---|
| Stable prefix | Prefer one canonical copy of long text in skills; rules point — avoid churn in alwaysApply: true bodies. |
| Lean always-on | Rarely needed norms → alwaysApply: false + globs, or load skill on demand. |
| Indirection | Agents/rules cite skill path; do not paste full playbooks into rules or agent bodies. |
| Task payloads | Parent Task prompts: goal, constraints, approved_plan_path, phase/shard id, touches[], correlation ids — not full rules or transcripts. |
| Reads | Batch parallel Read where independent; one read per file per turn unless content changed; lookup-first for brain-memory-kb. |
| Subagent traffic | Children: caveman-ultra + single YAML envelope per templates/subagent-response.yml.tmpl; parents: do not paste raw child YAML to user. |
Non-claim: Exact cache eligibility is vendor-defined; token reduction is always a win.
Load ~/ai-brain/org/global/config/memory-demotion.yml (via brain-memory-kb) for query ladder L0–L3 caps and demotion thresholds — do not duplicate full policy text here.
task-orchestrationbrain-memory-kbvp-research.mdc (single broker); never duplicate broker bullets elsewheresubagent-response-protocol + templates/subagent-response.yml.tmpltoken_stats: agent-observabilityalwaysApply rule bodies into chat to “teach” a subagentTask (anti-dup hook is fail-closed)