File size: 9,574 Bytes
3fbbaab
 
 
 
 
 
 
 
 
 
 
 
 
21ff762
3fbbaab
 
 
cafa909
 
 
 
 
3fbbaab
e965652
 
 
 
 
3fbbaab
 
 
 
 
 
 
 
 
 
e965652
 
 
 
 
 
21ff762
 
 
e965652
 
 
3fbbaab
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
916f576
3fbbaab
 
 
 
 
916f576
3fbbaab
 
 
 
 
e965652
3fbbaab
 
 
 
 
 
 
e965652
 
 
3fbbaab
e965652
 
 
 
 
 
 
 
 
 
 
 
 
3fbbaab
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e965652
 
 
 
 
 
 
3fbbaab
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
{
  "_comment": "Alive Skill System configuration. Edit this file to customize paths and thresholds.",
  "_config_path": "~/.claude/skill-system-config.json",

  "paths": {
    "claude_dir": "~/.claude",
    "wiki_dir": "~/.claude/skill-wiki",
    "skills_dir": "~/.claude/skills",
    "agents_dir": "~/.claude/agents",
    "skill_manifest": "~/.claude/skill-manifest.json",
    "intent_log": "~/.claude/intent-log.jsonl",
    "pending_skills": "~/.claude/pending-skills.json",
    "skill_registry": "~/.claude/skill-registry.json",
    "stack_profile_tmp": "~/.claude/skill-stack-profile.json",
    "catalog": "~/.claude/skill-wiki/catalog.md"
  },

  "knowledge": {
    "_comment": "User knowledge-source policy. shipped = use ctx's shipped graph/wiki; local = private/local wiki only; enriched = start from shipped ctx knowledge and add user/project content.",
    "mode": "shipped"
  },

  "resolver": {
    "_comment": "Resolver + recommendation tuning. max_skills is the HARD ceiling for a single resolve call (anti-budget-blowout). recommendation_top_k is the hard cap shown to users as an execution bundle — a mixed recommendation spanning skills + agents + MCPs ordered by graph score. Harnesses use the separate harness catalog flow. Keep top_k small (default/max 5) so users aren't overwhelmed by options; recommendation_min_normalized_score drops weak graph matches instead of recommending at all cost.",
    "max_skills": 15,
    "recommendation_top_k": 5,
    "recommendation_min_normalized_score": 0.30,
    "intent_boost_per_signal": 5,
    "intent_boost_max": 15,
    "staleness_penalty": -8,
    "meta_skills": ["skill-router", "file-reading"]
  },

  "context_monitor": {
    "unmatched_signal_threshold": 3,
    "manifest_stale_minutes": 60
  },

  "usage_tracker": {
    "stale_threshold_sessions": 30,
    "keep_log_days": 5
  },

  "harness": {
    "_comment": "Harness catalog recommendation tuning for custom/API/local model users. Harnesses are separate from the execution bundle and are only recommended when their absolute fit score is at or above recommendation_min_fit_score. The normalized floor is kept for older user configs and display compatibility.",
    "recommendation_min_normalized_score": 0.85,
    "recommendation_min_fit_score": 0.85
  },

  "skill_transformer": {
    "line_threshold": 180,
    "max_stage_lines": 40,
    "stage_count": 5
  },

  "skill_router": {
    "manifest_stale_minutes": 60,
    "manifest_max_age_hours": 24
  },

  "extra_skill_dirs": [],

  "intake": {
    "enabled": true,
    "dup_threshold": 0.93,
    "near_dup_threshold": 0.80,
    "min_neighbors": 0,
    "min_neighbor_score": 0.30,
    "min_body_chars": 120,
    "cache_root": "~/.claude/skills/_embeddings",
    "embedding": {
      "backend": "sentence-transformers",
      "model": null,
      "base_url": null,
      "allow_remote": false
    }
  },

  "graph": {
    "_comment": "Knowledge-graph edge weighting for wiki_graphify. Base edges between entity pages (skill ↔ agent ↔ mcp-server ↔ harness, any combination) come from semantic similarity, explicit tags, slug tokens, source overlap, or direct wikilinks. Semantic/tag/token weights must sum to 1.0. min_edge_weight drops weak final blended edges after scoring; default 0.03 is the calibrated shipped graph floor. Edge boosts then add explainable rank signals (direct links, source overlap, Adamic-Adar, type affinity, usage, quality) without creating edges by themselves. Override any key in your user config (~/.claude/skill-system-config.json).",
    "edge_weights": {
      "semantic":    0.70,
      "tags":        0.15,
      "slug_tokens": 0.15
    },
    "min_edge_weight": 0.03,
    "semantic": {
      "_comment": "Semantic edge construction uses two thresholds so you can tune display strictness WITHOUT paying to rebuild the graph. build_floor is the low bar for inclusion (all pairs at or above this survive to graph.json); min_cosine is the query/display filter that consumers (resolve, visualize, recommend) apply at read time. build_floor must be <= min_cosine. Rebuilding is only required when you change top_k, build_floor, or the entity body text that drives the cached embeddings — min_cosine alone is rebuild-free.",
      "top_k": 20,
      "build_floor": 0.50,
      "min_cosine": 0.80,
      "batch_size": 512,
      "cache_dir": "~/.claude/skill-wiki/.embedding-cache/graph"
    },
    "tag_edges": {
      "_comment": "Tag edge construction. dense_tag_threshold skips over-broad tags shared by >N nodes (a tag on 1000 nodes would create 500k noise edges of weight 1). Use Infinity to disable. shared_tag_saturation is the number of overlapping tags at which tag_sim saturates to 1.0 — with saturation=5, two pages sharing 5+ tags contribute tag_sim=1.0; sharing 2 contributes 0.4.",
      "dense_tag_threshold": 500,
      "shared_tag_saturation": 5
    },
    "token_edges": {
      "_comment": "Slug-token edge construction. Mirrors tag_edges but for the pseudo-tags synthesized from slugs. Keep dense_token_threshold intentionally low: after 90K+ hydrated Skills.sh bodies, broad slug words create millions of low-signal token-only edges. Semantic top-K and explicit tags are the primary graph signals; slug tokens are only a sparse fallback.",
      "dense_token_threshold": 30,
      "shared_token_saturation": 3
    },
    "source_edges": {
      "_comment": "Source-overlap edge construction. Pages with the same repo/source/homepage/detail/package URL can connect even when their tags differ. Buckets above dense_source_threshold are skipped so broad catalog URLs cannot create noise edges.",
      "dense_source_threshold": 50
    },
    "edge_boosts": {
      "_comment": "Additive edge boosts. These never create edges alone; they only strengthen pairs already justified by semantic, tags, slug tokens, source overlap, or direct wikilinks. Set a value to 0 to disable that signal.",
      "direct_link": 0.10,
      "source_overlap": 0.05,
      "adamic_adar": 0.04,
      "type_affinity": 0.03,
      "usage": 0.02,
      "quality": 0.02
    }
  },

  "mcp_quality": {
    "_comment": "Phase 4 MCP scorer. Weights must sum to 1.0; signals are popularity, freshness, structural, graph, trust, runtime. Override per-key in your user config (~/.claude/skill-system-config.json).",
    "weights": {
      "popularity": 0.30,
      "freshness":  0.20,
      "structural": 0.15,
      "graph":      0.15,
      "trust":      0.10,
      "runtime":    0.10
    },
    "grade_thresholds": {
      "A": 0.80,
      "B": 0.60,
      "C": 0.40
    },
    "star_saturation": 1000,
    "freshness_half_life_days": 90.0,
    "graph_degree_saturation": 20,
    "paths": {
      "sidecar_dir": "~/.claude/skill-quality/mcp"
    }
  },

  "babysitter": {
    "plugin_root": "",
    "runs_dir": ".a5c/runs",
    "sdk_version": "latest"
  },

  "backup": {
    "_comment": "Change-triggered snapshot config. Paths under top_files are relative to ~/.claude. Trees copy directories. memory_glob walks ~/.claude/projects/*/memory.",
    "snapshot_dir": "~/.claude/backups",
    "name_format": "{timestamp}",
    "timestamp_format": "%Y%m%dT%H%M%SZ",
    "scope": "full",
    "max_file_bytes": 5242880,
    "top_files": [
      "settings.json",
      "skill-manifest.json",
      "pending-skills.json",
      "CLAUDE.md",
      "AGENTS.md",
      "user-profile.json",
      "skill-system-config.json",
      "skill-registry.json"
    ],
    "trees": [
      {"src": "agents", "dest": "agents"},
      {"src": "skills", "dest": "skills"}
    ],
    "memory_glob": true,
    "excludes": [],
    "retention": {
      "keep_latest": 50,
      "keep_daily": 14
    }
  },

  "quality": {
    "_comment": "Post-install quality scorer. 'weights' applies to skills (telemetry-heavy — skills are auto-loaded and usage frequency is the strongest post-install signal). 'agent_weights' applies to agents (structure-heavy — agents are invoked deliberately and rarely, so telemetry is a weaker quality signal). Both must sum to 1.0 and contain exactly: telemetry, intake, graph, routing. Grade thresholds must satisfy 0 <= C <= B <= A <= 1. See src/skill_quality.py.",
    "weights": {
      "telemetry": 0.40,
      "intake": 0.20,
      "graph": 0.25,
      "routing": 0.15
    },
    "agent_weights": {
      "telemetry": 0.15,
      "intake": 0.30,
      "graph": 0.35,
      "routing": 0.20
    },
    "grade_thresholds": {
      "A": 0.80,
      "B": 0.60,
      "C": 0.40
    },
    "stale_threshold_days": 30.0,
    "recent_window_days": 14.0,
    "min_body_chars": 120,
    "paths": {
      "sidecar_dir": "~/.claude/skill-quality",
      "router_trace": "~/.claude/router-trace.jsonl"
    },
    "lifecycle": {
      "_comment": "Asymmetric lifecycle gates. Demotions happen automatically from a D-streak; archive requires the skill to sit in _demoted long enough; delete requires typed-slug confirmation. See src/ctx_lifecycle.py.",
      "archive_threshold_days": 14.0,
      "delete_threshold_days": 60.0,
      "consecutive_d_to_demote": 2,
      "demoted_subdir": "_demoted",
      "archive_subdir": "_archive",
      "history_max": 20
    },
    "dashboard": {
      "_comment": "KPI dashboard rendering knobs. See src/kpi_dashboard.py.",
      "default_top_n": 10,
      "report_path": "~/.claude/skill-quality/kpi.md"
    }
  }
}