[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M71C1Y9KBJT209KBXBTE022F":3,"skill-same:01M71C1Y9KBJT209KBXBTE022F":73},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":39,"linkOnly":40,"markdownContent":41,"markdownBody":42,"frontMatter":43,"refName":52,"commitSha":53,"skillDirectory":54,"forksCount":55,"repositoryHtmlUrl":56,"repositoryHomepageUrl":6,"ownerHtmlUrl":57,"ownerBlogUrl":6,"ownerDescription":58,"updatedAt":59,"install":60,"resources":72},"01M71C1Y9KBJT209KBXBTE022F","compact-optimizer",null,"Construct the minimum-viable \u002Fcompact hint using selective context compression (self-information scoring, perplexity-based pruning, attention-sink preservation). Use when about to call \u002Fcompact, when context is nearing capacity, or when you need to preserve session-critical state through a compression boundary.","nexuslinkproductions\u002Fyuri-os","nexuslinkproductions","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F273246263?v=4","01M718YX9NZRCW50B59B8ZVDKN","Marcel Spatz","01M718YXAQ7530F4B1VKN6SFQV",2,"MIT","https:\u002F\u002Fgithub.com\u002Fnexuslinkproductions\u002Fyuri-os\u002Fblob\u002Fmain\u002Fskills\u002Fcompact-optimizer\u002FSKILL.md",[],[19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38],"optimizer","compact optimizer","compact","true","long","session","attention","harness","running","perplexity","context","compress","hint","self","scoring","you","when","critical","information","context compression","2026-11-10 10:12:30",false,"---\nname: compact-optimizer\ndisable-model-invocation: true\ndescription: \"Construct the minimum-viable \u002Fcompact hint using selective context compression (self-information scoring, perplexity-based pruning, attention-sink preservation). Use when about to call \u002Fcompact, when context is nearing capacity, or when you need to preserve session-critical state through a compression boundary.\"\ntriggers:\n  - \"\u002Fcompact-hint\"\n  - \"context is getting long\"\n  - \"running out of context\"\n  - \"compact before\"\n  - \"compress the context\"\nscope: harness\ninvocation: ability\n---\n\n# Compact Optimizer\n\nOne job: build the smallest possible hint that tells \u002Fcompact exactly what must survive.\n\n## Research Grounding\n\n| Paper | Core Rule |\n|-------|-----------|\n| Selective Context (EMNLP 2023) | Drop what is reconstructable from general knowledge. Preserve what is unique to this session. |\n| LLMLingua (EMNLP 2023) | Compress reasoning\u002Fexploration 60–80%. Compress instructions\u002Fconstraints ≤20%. |\n| H2O MIT (NeurIPS 2023) | Always retain the last user correction + last known file state — they are attention sinks. |\n\n---\n\n## Survival Priority Hierarchy\n\nRank 1 — **Never drop** (session-unique, irreplaceable):\n- Active branch name and working directory\n- Files modified in this session (paths + what changed)\n- User corrections, explicit rejections, aversion decisions\n- Definition of done \u002F success criteria\n- Last error message or test failure state\n- Approved plan steps and which are complete\n\nRank 2 — **Compress heavily** (preserve meaning, strip words):\n- Current task description (1 sentence max)\n- Key constraints stated by user\n- Architecture decisions made\n\nRank 3 — **Drop entirely** (reconstructable, low self-information):\n- Exploration text (\"let me look at...\", \"I'll check...\")\n- Preambles and acknowledgments\n- Reasoning narration that led to a decision already made\n- Redundant tool call descriptions (keep results, drop scaffolding)\n- Any content the model could reproduce from the codebase alone\n\n---\n\n## Compact Hint Template\n\n```\n\u002Fcompact Preserve: [branch\u002Fdir] | Files touched: [path1, path2] | Last correction: [exact user correction] | Current task: [1 sentence] | Done criteria: [condition] | Status: [step N of M complete, next: X]\n```\n\n**Rules for the hint itself:**\n- Max 3 sentences. No preamble.\n- Lead with the file state — it is always the highest-attention-sink.\n- Include the last user correction verbatim if one exists.\n- Name the next concrete action.\n- Never include reasoning or exploration in the hint.\n\n---\n\n## Per-Phase Templates\n\n**Research phase**\n```\n\u002Fcompact Preserve: research goal=[goal], key finding=[finding], next: [what to research next]\n```\n\n**Implementation phase** (most common)\n```\n\u002Fcompact Preserve: branch=[branch], files=[paths], constraint=[constraint], last correction=[correction], next step=[action], done when=[criteria]\n```\n\n**Review \u002F debug phase**\n```\n\u002Fcompact Preserve: error=[error msg], file=[path:line], hypothesis=[current], last tried=[approach], next: [next attempt]\n```\n\n---\n\n## When to Call \u002Fcompact\n\n- Context bar >60% (tokenmaxxing mode) or >65% (standard mode): run \u002Fcompact with hint — thresholds match the tier table in pre-tool-use.js getTier()\n- Phase change (research→implement, implement→review): always \u002Fcompact with phase-transition hint\n- After a user correction: \u002Fcompact immediately, include correction verbatim\n- Never compact mid-tool-use or mid-plan without noting the interruption point\n\n---\n\n## Anti-patterns (never do these)\n\n- `\u002Fcompact` with no hint → generic summary, drops constraints\n- Hint longer than 3 sentences → defeats compression purpose\n- Including reasoning in the hint → low self-information, wastes hint budget\n- Compacting after a failed approach without writing an Aversion Memory node first\n\n---\n\n## Aversion Memory Gate\n\nPer the YURI Aversion Memory Protocol: if you are compacting after a **failed branch**, first write the failure reason to an Aversion Memory node, THEN compact. The compact hint should reference the aversion: `last aversion: [reason]`.\n\n## Token Budget Policy\n\n- Trigger \u002Fcompact at 60%+ (tokenmaxxing) \u002F 65%+ (standard) context. A long transcript (~40k lines) is a guideline for manual \u002Fcompact — no hook enforces a transcript-line limit.\n- Dirty repo + mid-sprint: capture scoped `git status --short` marker before compacting.\n- Lane output arriving in main context: summarize to marker-only before injection.\n- After \u002Fcompact, re-read 2–3 key touched files to verify compact summary accuracy.\n\n## Evidence Pack Shape\n\nCanonical compact payload for DeepSeek reinforcement and lane handoff:\n\n```\nTASK: \u003Cone sentence>\nCONTEXT_PACK:\n- fact1\n- fact2  (max 5 facts)\nEVIDENCE:\n- ref\u002Fsnippet1\n- ref\u002Fsnippet2  (max 3 refs)\nBLOCKERS: \u003Cnone | specific>\nQUESTION: \u003Cspecific question or \"proceed\">\nOUTPUT_CAP: \u003C80 lines research | 120 lines final report>\n```\n\nNever inject raw command dumps, full file contents, or unfiltered grep output into a compact payload.\n\n## Reversible Path (ccr-compress)\n\nThis skill builds a ONE-WAY hint — the tokens it tells `\u002Fcompact` to drop are gone. When a section is too valuable to lose-permanently but too heavy to keep inline, route it through the reversible compaction mechanism instead:\n\n- `_SYSTEM\u002FScripts\u002Fccr-compress.mjs` — `compress(payload)` caches the BYTE-EXACT original under a content hash (sibling dir `_SYSTEM\u002Fstate\u002Fccr-cache\u002F`, TTL-pruned, gitignored) and injects a retrieval **sentinel** (`⟪CCR:\u003Chash>:\u003Ctype>:\u003Cbytes>⟫`) in its place. `retrieve(hash)` pulls the original back. The cache is the undo button — that is what makes dropping a section safe.\n- **Content-typed + honest about loss:** `json`\u002F`code`\u002F`prose` routing. Structural elision is **lossless** (round-trips byte-exact). Semantic elision is marked `lossy:true` — the cache still restores the source, but the inline placeholder is not self-reconstructable. Never claim semantic inline is reversible.\n- **Cache-prefix conscience:** `_SYSTEM\u002FScripts\u002Fcache-prefix-scan.mjs` is a WARN-ONLY detector — it flags volatile tokens (UUID \u002F ISO-timestamp \u002F JWT \u002F long-hex \u002F epoch-ms) leaking into the KV-cache-hot prefix and NEVER mutates the prefix. Pairs with the \"Token Caching Shape\" rule in CLAUDE.md (keep volatile ids out of the stable preamble). Lane label: `04CP_CACHE_PREFIX_SCAN_X_COMMITTED`.\n- **`--self` is hardened + diagnose-only:** reads ONLY `global.md` + `MEMORY.md`, never invokes `brain-inject` (which reads the deny-listed `.claude\u002Fstate\u002Fcortex-state.json`). It reports self-context headroom; it does not repair context.\n\nReach for ccr-compress when the answer to \"can I get this back if I need it?\" must be yes. Use the plain `\u002Fcompact` hint above when the dropped content is genuinely reconstructable from the codebase.\n\n## Session Notes\n\n### 2026-06-16\n- session: 7m | peak ctx: 0% | compacts: 0\n- tools: Bash×9, Read×6, Edit×5, WebSearch×3, Write×1, Skill×1, ToolSearch×1\n- corrections: none\n- errors: none\n\n### 2026-06-15\n- session: 11m | peak ctx: 0% | compacts: 0\n- tools: Bash×19, Read×2, Edit×2, Skill×1, Write×1\n- corrections: Base directory for this skill: \u002FUsers\u002Fmarcelspatz\u002F.claude\u002Fskills\u002Fcross-reference-navigation\n\n# Cross-Reference Navigation (XREF)\n\nThe GROUND step of the work loop, made reflexive. One question asked a | Base directory for this skill: \u002FUsers\u002Fmarcelspatz\u002F.claude\u002Fskills\u002Fquantum-hypothesis-simulation\n\n# Quantum Hypothesis Simulation\n\nThe quantum-probability layer for YURI's claim\u002Fpulse machinery. It mode | Base directory for this skill: \u002FUsers\u002Fmarcelspatz\u002F.claude\u002Fskills\u002Fcross-reference-navigation\n\n# Cross-Reference Navigation (XREF)\n\nThe GROUND step of the work loop, made reflexive. One question asked a\n- errors: none\n\n### 2026-06-14\n- session: 101m | peak ctx: 0% | compacts: 0\n- tools: Bash×455, Read×142, Edit×74, Write×44, WebSearch×31, WebFetch×19, StructuredOutput×17, ToolSearch×8, TodoWrite×5, Workflow×3, AskUserQuestion×2, Skill×2, TaskStop×1\n- corrections: Base directory for this skill: \u002FUsers\u002Fmarcelspatz\u002F.claude\u002Fskills\u002Fcross-reference-navigation\n\n# Cross-Reference Navigation (XREF)\n\nThe GROUND step of the work loop, made reflexive. One question asked a | Base directory for this skill: \u002FUsers\u002Fmarcelspatz\u002F.claude\u002Fskills\u002Fquantum-hypothesis-simulation\n\n# Quantum Hypothesis Simulation\n\nThe quantum-probability layer for YURI's claim\u002Fpulse machinery. It mode | Base directory for this skill: \u002FUsers\u002Fmarcelspatz\u002F.claude\u002Fskills\u002Fcross-reference-navigation\n\n# Cross-Reference Navigation (XREF)\n\nThe GROUND step of the work loop, made reflexive. One question asked a\n- errors: none\n\n### 2026-06-13\n- session: 9m | peak ctx: 0% | compacts: 0\n- tools: Bash×45, Write×16, Edit×13, Read×10, TaskStop×1, Skill×1\n- corrections: rick i have a fun little task for you. I will be giving you the task of going through trending repos on github, scanning them, compare yuri to those, see what we can adopt and rebuild better in yuri u\n- errors: none\n\n### 2026-06-12\n- session: 97m | peak ctx: 100% | compacts: 3\n- tools: Edit×67, Bash×47, Read×42, Write×5, WebFetch×3, Skill×3, TaskUpdate×2, WebSearch×2, TaskCreate×1, EnterPlanMode×1\n- corrections: none\n- errors: none\n\n### 2026-06-11\n- session: 70m | peak ctx: 100% | compacts: 3\n- tools: Read×34, Edit×30, Bash×19, Write×5, WebFetch×3, Skill×3, TaskUpdate×2, WebSearch×2, TaskCreate×1, EnterPlanMode×1\n- corrections: none\n- errors: none\n\n### 2026-06-10\n- session: 847m | peak ctx: 62% | compacts: 1\n- tools: Bash×776, Read×365, Edit×55, Write×27, StructuredOutput×12, TodoWrite×11, WebFetch×9, Agent×7, Workflow×3, ToolSearch×2\n- corrections: none\n- errors: none\n\n### 2026-06-04\n- session: 90m | peak ctx: 0% | compacts: 0\n- tools: Bash×452, Read×180, StructuredOutput×34, WebSearch×19, Edit×13, Write×10, TodoWrite×8, WebFetch×7, ToolSearch×6, Workflow×4, mcp×1\n- corrections: none\n- errors: none\n\n### 2026-06-03\n- session: 147m | peak ctx: 0% | compacts: 0\n- tools: Bash×798, Read×626, Write×163, StructuredOutput×140, WebSearch×84, Edit×75, ToolSearch×39, WebFetch×16, Workflow×7, AskUserQuestion×1, Agent×1\n- corrections: none\n- errors: none\n\n### 2026-05-02\n- session: 4m | peak ctx: 14% | compacts: 0\n- tools: Bash×16, Read×4, Edit×4, Skill×1\n- corrections: none\n- errors: none\n\n### 2026-04-27\n- session: 1m | peak ctx: 43% | compacts: 0\n- tools: Bash×15, Read×12\n- corrections: none\n- errors: none\n\n### 2026-04-26\n- session: 7m | peak ctx: 0% | compacts: 0\n- tools: Bash×15, Read×9, Write×4, Agent×1, ToolSearch×1, ExitPlanMode×1, Edit×1\n- corrections: none\n- errors: none\n\n### 2026-04-25\n- session: 0m | peak ctx: 14% | compacts: 0\n- tools: Read×9, Bash×4, Write×2, Edit×2\n- corrections: none\n- errors: none\n\n### 2026-04-24\n- session: 0m | peak ctx: 21% | compacts: 0 (mid-session)\n- tools: Bash×3, Read×1\n- corrections: none\n- errors: exit code 1: command not found\n","# Compact Optimizer\n\nOne job: build the smallest possible hint that tells \u002Fcompact exactly what must survive.\n\n## Research Grounding\n\n| Paper | Core Rule |\n|-------|-----------|\n| Selective Context (EMNLP 2023) | Drop what is reconstructable from general knowledge. Preserve what is unique to this session. |\n| LLMLingua (EMNLP 2023) | Compress reasoning\u002Fexploration 60–80%. Compress instructions\u002Fconstraints ≤20%. |\n| H2O MIT (NeurIPS 2023) | Always retain the last user correction + last known file state — they are attention sinks. |\n\n---\n\n## Survival Priority Hierarchy\n\nRank 1 — **Never drop** (session-unique, irreplaceable):\n- Active branch name and working directory\n- Files modified in this session (paths + what changed)\n- User corrections, explicit rejections, aversion decisions\n- Definition of done \u002F success criteria\n- Last error message or test failure state\n- Approved plan steps and which are complete\n\nRank 2 — **Compress heavily** (preserve meaning, strip words):\n- Current task description (1 sentence max)\n- Key constraints stated by user\n- Architecture decisions made\n\nRank 3 — **Drop entirely** (reconstructable, low self-information):\n- Exploration text (\"let me look at...\", \"I'll check...\")\n- Preambles and acknowledgments\n- Reasoning narration that led to a decision already made\n- Redundant tool call descriptions (keep results, drop scaffolding)\n- Any content the model could reproduce from the codebase alone\n\n---\n\n## Compact Hint Template\n\n```\n\u002Fcompact Preserve: [branch\u002Fdir] | Files touched: [path1, path2] | Last correction: [exact user correction] | Current task: [1 sentence] | Done criteria: [condition] | Status: [step N of M complete, next: X]\n```\n\n**Rules for the hint itself:**\n- Max 3 sentences. No preamble.\n- Lead with the file state — it is always the highest-attention-sink.\n- Include the last user correction verbatim if one exists.\n- Name the next concrete action.\n- Never include reasoning or exploration in the hint.\n\n---\n\n## Per-Phase Templates\n\n**Research phase**\n```\n\u002Fcompact Preserve: research goal=[goal], key finding=[finding], next: [what to research next]\n```\n\n**Implementation phase** (most common)\n```\n\u002Fcompact Preserve: branch=[branch], files=[paths], constraint=[constraint], last correction=[correction], next step=[action], done when=[criteria]\n```\n\n**Review \u002F debug phase**\n```\n\u002Fcompact Preserve: error=[error msg], file=[path:line], hypothesis=[current], last tried=[approach], next: [next attempt]\n```\n\n---\n\n## When to Call \u002Fcompact\n\n- Context bar >60% (tokenmaxxing mode) or >65% (standard mode): run \u002Fcompact with hint — thresholds match the tier table in pre-tool-use.js getTier()\n- Phase change (research→implement, implement→review): always \u002Fcompact with phase-transition hint\n- After a user correction: \u002Fcompact immediately, include correction verbatim\n- Never compact mid-tool-use or mid-plan without noting the interruption point\n\n---\n\n## Anti-patterns (never do these)\n\n- `\u002Fcompact` with no hint → generic summary, drops constraints\n- Hint longer than 3 sentences → defeats compression purpose\n- Including reasoning in the hint → low self-information, wastes hint budget\n- Compacting after a failed approach without writing an Aversion Memory node first\n\n---\n\n## Aversion Memory Gate\n\nPer the YURI Aversion Memory Protocol: if you are compacting after a **failed branch**, first write the failure reason to an Aversion Memory node, THEN compact. The compact hint should reference the aversion: `last aversion: [reason]`.\n\n## Token Budget Policy\n\n- Trigger \u002Fcompact at 60%+ (tokenmaxxing) \u002F 65%+ (standard) context. A long transcript (~40k lines) is a guideline for manual \u002Fcompact — no hook enforces a transcript-line limit.\n- Dirty repo + mid-sprint: capture scoped `git status --short` marker before compacting.\n- Lane output arriving in main context: summarize to marker-only before injection.\n- After \u002Fcompact, re-read 2–3 key touched files to verify compact summary accuracy.\n\n## Evidence Pack Shape\n\nCanonical compact payload for DeepSeek reinforcement and lane handoff:\n\n```\nTASK: \u003Cone sentence>\nCONTEXT_PACK:\n- fact1\n- fact2  (max 5 facts)\nEVIDENCE:\n- ref\u002Fsnippet1\n- ref\u002Fsnippet2  (max 3 refs)\nBLOCKERS: \u003Cnone | specific>\nQUESTION: \u003Cspecific question or \"proceed\">\nOUTPUT_CAP: \u003C80 lines research | 120 lines final report>\n```\n\nNever inject raw command dumps, full file contents, or unfiltered grep output into a compact payload.\n\n## Reversible Path (ccr-compress)\n\nThis skill builds a ONE-WAY hint — the tokens it tells `\u002Fcompact` to drop are gone. When a section is too valuable to lose-permanently but too heavy to keep inline, route it through the reversible compaction mechanism instead:\n\n- `_SYSTEM\u002FScripts\u002Fccr-compress.mjs` — `compress(payload)` caches the BYTE-EXACT original under a content hash (sibling dir `_SYSTEM\u002Fstate\u002Fccr-cache\u002F`, TTL-pruned, gitignored) and injects a retrieval **sentinel** (`⟪CCR:\u003Chash>:\u003Ctype>:\u003Cbytes>⟫`) in its place. `retrieve(hash)` pulls the original back. The cache is the undo button — that is what makes dropping a section safe.\n- **Content-typed + honest about loss:** `json`\u002F`code`\u002F`prose` routing. Structural elision is **lossless** (round-trips byte-exact). Semantic elision is marked `lossy:true` — the cache still restores the source, but the inline placeholder is not self-reconstructable. Never claim semantic inline is reversible.\n- **Cache-prefix conscience:** `_SYSTEM\u002FScripts\u002Fcache-prefix-scan.mjs` is a WARN-ONLY detector — it flags volatile tokens (UUID \u002F ISO-timestamp \u002F JWT \u002F long-hex \u002F epoch-ms) leaking into the KV-cache-hot prefix and NEVER mutates the prefix. Pairs with the \"Token Caching Shape\" rule in CLAUDE.md (keep volatile ids out of the stable preamble). Lane label: `04CP_CACHE_PREFIX_SCAN_X_COMMITTED`.\n- **`--self` is hardened + diagnose-only:** reads ONLY `global.md` + `MEMORY.md`, never invokes `brain-inject` (which reads the deny-listed `.claude\u002Fstate\u002Fcortex-state.json`). It reports self-context headroom; it does not repair context.\n\nReach for ccr-compress when the answer to \"can I get this back if I need it?\" must be yes. Use the plain `\u002Fcompact` hint above when the dropped content is genuinely reconstructable from the codebase.\n\n## Session Notes\n\n### 2026-06-16\n- session: 7m | peak ctx: 0% | compacts: 0\n- tools: Bash×9, Read×6, Edit×5, WebSearch×3, Write×1, Skill×1, ToolSearch×1\n- corrections: none\n- errors: none\n\n### 2026-06-15\n- session: 11m | peak ctx: 0% | compacts: 0\n- tools: Bash×19, Read×2, Edit×2, Skill×1, Write×1\n- corrections: Base directory for this skill: \u002FUsers\u002Fmarcelspatz\u002F.claude\u002Fskills\u002Fcross-reference-navigation\n\n# Cross-Reference Navigation (XREF)\n\nThe GROUND step of the work loop, made reflexive. One question asked a | Base directory for this skill: \u002FUsers\u002Fmarcelspatz\u002F.claude\u002Fskills\u002Fquantum-hypothesis-simulation\n\n# Quantum Hypothesis Simulation\n\nThe quantum-probability layer for YURI's claim\u002Fpulse machinery. It mode | Base directory for this skill: \u002FUsers\u002Fmarcelspatz\u002F.claude\u002Fskills\u002Fcross-reference-navigation\n\n# Cross-Reference Navigation (XREF)\n\nThe GROUND step of the work loop, made reflexive. One question asked a\n- errors: none\n\n### 2026-06-14\n- session: 101m | peak ctx: 0% | compacts: 0\n- tools: Bash×455, Read×142, Edit×74, Write×44, WebSearch×31, WebFetch×19, StructuredOutput×17, ToolSearch×8, TodoWrite×5, Workflow×3, AskUserQuestion×2, Skill×2, TaskStop×1\n- corrections: Base directory for this skill: \u002FUsers\u002Fmarcelspatz\u002F.claude\u002Fskills\u002Fcross-reference-navigation\n\n# Cross-Reference Navigation (XREF)\n\nThe GROUND step of the work loop, made reflexive. One question asked a | Base directory for this skill: \u002FUsers\u002Fmarcelspatz\u002F.claude\u002Fskills\u002Fquantum-hypothesis-simulation\n\n# Quantum Hypothesis Simulation\n\nThe quantum-probability layer for YURI's claim\u002Fpulse machinery. It mode | Base directory for this skill: \u002FUsers\u002Fmarcelspatz\u002F.claude\u002Fskills\u002Fcross-reference-navigation\n\n# Cross-Reference Navigation (XREF)\n\nThe GROUND step of the work loop, made reflexive. One question asked a\n- errors: none\n\n### 2026-06-13\n- session: 9m | peak ctx: 0% | compacts: 0\n- tools: Bash×45, Write×16, Edit×13, Read×10, TaskStop×1, Skill×1\n- corrections: rick i have a fun little task for you. I will be giving you the task of going through trending repos on github, scanning them, compare yuri to those, see what we can adopt and rebuild better in yuri u\n- errors: none\n\n### 2026-06-12\n- session: 97m | peak ctx: 100% | compacts: 3\n- tools: Edit×67, Bash×47, Read×42, Write×5, WebFetch×3, Skill×3, TaskUpdate×2, WebSearch×2, TaskCreate×1, EnterPlanMode×1\n- corrections: none\n- errors: none\n\n### 2026-06-11\n- session: 70m | peak ctx: 100% | compacts: 3\n- tools: Read×34, Edit×30, Bash×19, Write×5, WebFetch×3, Skill×3, TaskUpdate×2, WebSearch×2, TaskCreate×1, EnterPlanMode×1\n- corrections: none\n- errors: none\n\n### 2026-06-10\n- session: 847m | peak ctx: 62% | compacts: 1\n- tools: Bash×776, Read×365, Edit×55, Write×27, StructuredOutput×12, TodoWrite×11, WebFetch×9, Agent×7, Workflow×3, ToolSearch×2\n- corrections: none\n- errors: none\n\n### 2026-06-04\n- session: 90m | peak ctx: 0% | compacts: 0\n- tools: Bash×452, Read×180, StructuredOutput×34, WebSearch×19, Edit×13, Write×10, TodoWrite×8, WebFetch×7, ToolSearch×6, Workflow×4, mcp×1\n- corrections: none\n- errors: none\n\n### 2026-06-03\n- session: 147m | peak ctx: 0% | compacts: 0\n- tools: Bash×798, Read×626, Write×163, StructuredOutput×140, WebSearch×84, Edit×75, ToolSearch×39, WebFetch×16, Workflow×7, AskUserQuestion×1, Agent×1\n- corrections: none\n- errors: none\n\n### 2026-05-02\n- session: 4m | peak ctx: 14% | compacts: 0\n- tools: Bash×16, Read×4, Edit×4, Skill×1\n- corrections: none\n- errors: none\n\n### 2026-04-27\n- session: 1m | peak ctx: 43% | compacts: 0\n- tools: Bash×15, Read×12\n- corrections: none\n- errors: none\n\n### 2026-04-26\n- session: 7m | peak ctx: 0% | compacts: 0\n- tools: Bash×15, Read×9, Write×4, Agent×1, ToolSearch×1, ExitPlanMode×1, Edit×1\n- corrections: none\n- errors: none\n\n### 2026-04-25\n- session: 0m | peak ctx: 14% | compacts: 0\n- tools: Read×9, Bash×4, Write×2, Edit×2\n- corrections: none\n- errors: none\n\n### 2026-04-24\n- session: 0m | peak ctx: 21% | compacts: 0 (mid-session)\n- tools: Bash×3, Read×1\n- corrections: none\n- errors: exit code 1: command not found\n",{"name":5,"scope":26,"triggers":44,"invocation":50,"description":7,"disable-model-invocation":51},[45,46,47,48,49],"\u002Fcompact-hint","context is getting long","running out of context","compact before","compress the context","ability",true,"main","4c6d8d65f3d6096e2fc25b0845e481dda73c2794","skills\u002Fcompact-optimizer",0,"https:\u002F\u002Fgithub.com\u002Fnexuslinkproductions\u002Fyuri-os","https:\u002F\u002Fgithub.com\u002Fnexuslinkproductions","Building deterministic AI governance and continuity systems. Workflow architect + visual storyteller","2026-11-11 16:42:59",{"source":8,"skillName":5,"commands":61,"prompt":71,"zipAvailable":51},[62,65,68],{"tool":63,"command":64},"npx","npx skills add nexuslinkproductions\u002Fyuri-os --skill compact-optimizer",{"tool":66,"command":67},"bunx","bunx skills add nexuslinkproductions\u002Fyuri-os --skill compact-optimizer",{"tool":69,"command":70},"pnpm","pnpm dlx skills add nexuslinkproductions\u002Fyuri-os --skill compact-optimizer","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fnexuslinkproductions\u002Fyuri-os\u002Fblob\u002Fmain\u002Fskills\u002Fcompact-optimizer\u002FSKILL.md",[],[74,92,116,146,174,193],{"skillNo":75,"name":76,"nameZh":6,"description":77,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":78,"categories":79,"keywords":80,"publishedAt":39},"01M71CG1AME5WC0A79AE65WT0X","zoom-out","Tell the agent to zoom out and give broader context or a higher-level perspective. Use when you're unfamiliar with a section of code or need to understand how it fits into the bigger picture.","https:\u002F\u002Fgithub.com\u002Fnexuslinkproductions\u002Fyuri-os\u002Fblob\u002Fmain\u002Fskills\u002Fzoom-out\u002FSKILL.md",[],[81,82,83,26,84,85,86,22,87,88,89,35,34,90,91,29],"zoom out","zoom","with","how","section","re","understand","use","perspective","code","it",{"skillNo":93,"name":94,"nameZh":6,"description":95,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":96,"categories":97,"keywords":98,"publishedAt":39},"01M71CFY9ATD0BAXV6B14HFG7C","yuri-shura","6-perspective adversarial review for high-stakes turns (architecture decisions, refactor planning, deployment review) — fans out architect, adversary (7-vector attack), maintainer, ops, product, and security perspectives in parallel, then consolidates. Use when user says 'review this decision', '\u002Fshura', 'adversarial review', 'what could go wrong', 'architecture review', or 'deployment review'.","https:\u002F\u002Fgithub.com\u002Fnexuslinkproductions\u002Fyuri-os\u002Fblob\u002Fmain\u002Fskills\u002Fyuri-shura\u002FSKILL.md",[],[99,100,26,101,102,103,104,105,106,107,108,88,109,35,110,111,112,113,114,115],"yuri shura","parallel","architecture review","planning","for","deployment","vector","strategic review","decision","strategic","ops","workflow","decisions","architecture","review","perspectives","adversarial",{"skillNo":117,"name":118,"nameZh":6,"description":119,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":120,"categories":121,"keywords":127,"publishedAt":39},"01M71CFTTYW1HA2MNX38EV1GJ5","yuri-sales-intelligence","Unified sales and outreach skill generating cold outreach openers, social selling sequences, objection rebuttals, and ad creative for any paid platform — multi-persona aware, adapting tone and offer frame to buyer type. Use when user says 'write a cold email', 'outreach sequence', 'objection handling', 'ad copy', or 'sales pitch'.","https:\u002F\u002Fgithub.com\u002Fnexuslinkproductions\u002Fyuri-os\u002Fblob\u002Fmain\u002Fskills\u002Fyuri-sales-intelligence\u002FSKILL.md",[122],{"categoryNo":123,"categoryName":124,"taxonomyType":125,"taxonomyTypeDesc":126},"01JBSEEDCAT000000000000005","邮件与沟通","SCENARIO","使用场景",[128,129,130,131,132,133,134,135,136,35,137,138,139,140,141,142,143,103,144,145],"intelligence","sales intelligence","yuri sales intelligence","sales","cold","paid","social","persona","pitch","cold email","ad creative","write","sequences","sequence","platform","email","cold outreach","frame",{"skillNo":147,"name":148,"nameZh":6,"description":149,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":150,"categories":151,"keywords":160,"publishedAt":39},"01M71CFQMBFTFADJNGP0Z2WCYK","yuri-code-intelligence","Unified code quality intelligence layer that diagnoses code smells (Long Method, Feature Envy, design pattern issues), prescribes targeted Fowler refactoring techniques, and reviews code for security, performance, and clarity. Use when user says 'review this code', 'refactor this', 'what's wrong with this method', 'code smells', or 'clean this up'.","https:\u002F\u002Fgithub.com\u002Fnexuslinkproductions\u002Fyuri-os\u002Fblob\u002Fmain\u002Fskills\u002Fyuri-code-intelligence\u002FSKILL.md",[152,155],{"categoryNo":153,"categoryName":154,"taxonomyType":125,"taxonomyTypeDesc":126},"01JBSEEDCAT000000000000007","代码审查",{"categoryNo":156,"categoryName":157,"taxonomyType":158,"taxonomyTypeDesc":159},"01JBSEEDCAT000000000000030","开发工程师","PROFESSION","适用职业",[161,90,162,128,26,163,164,165,103,166,167,23,168,169,170,171,172,173,88,35],"yuri code intelligence","code intelligence","reviews","code quality","feature","performance","code review","up","issues","clean","techniques","improve","design",{"skillNo":175,"name":176,"nameZh":6,"description":177,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":178,"categories":179,"keywords":180,"publishedAt":39},"01M71CFP32RT6GD3GKNHCJP8VP","yuri-adw","Invoke when user asks to run the ADW loop \u002F agentic engineering loop \u002F software factory workflow.","https:\u002F\u002Fgithub.com\u002Fnexuslinkproductions\u002Fyuri-os\u002Fblob\u002Fmain\u002Fskills\u002Fyuri-adw\u002FSKILL.md",[],[181,182,183,184,185,186,187,188,189,35,190,110,191,192],"yuri adw","run","engineering loop","software factory","software","agentic engineering","factory","loop","invoke","engineering","agentic","user",{"skillNo":194,"name":195,"nameZh":6,"description":196,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":197,"categories":198,"keywords":211,"publishedAt":39},"01M71CFK4VPEKY6A73237P7RSN","writing-skills","Use when creating new skills, editing existing skills, or verifying skills work before deployment","https:\u002F\u002Fgithub.com\u002Fnexuslinkproductions\u002Fyuri-os\u002Fblob\u002Fmain\u002Fskills\u002Fwriting-skills\u002FSKILL.md",[199,202,205,208],{"categoryNo":200,"categoryName":201,"taxonomyType":125,"taxonomyTypeDesc":126},"01JBSEEDCAT000000000000008","测试与质量",{"categoryNo":203,"categoryName":204,"taxonomyType":125,"taxonomyTypeDesc":126},"01JBSEEDCAT000000000000018","写作与文案",{"categoryNo":206,"categoryName":207,"taxonomyType":125,"taxonomyTypeDesc":126},"01JBSEEDCAT000000000000023","Skill与智能体开发",{"categoryNo":209,"categoryName":210,"taxonomyType":158,"taxonomyTypeDesc":159},"01JBSEEDCAT000000000000036","内容创作者",[212,213,214,35,110,215,26,216,104,88,217],"writing","writing skills","new","work","editing","Existing"]