[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M79AP6AA8AXHD8V1FZKYARED":3,"skill-same:01M79AP6AA8AXHD8V1FZKYARED":51},{"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":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":29,"commitSha":30,"skillDirectory":31,"forksCount":32,"repositoryHtmlUrl":33,"repositoryHomepageUrl":34,"ownerHtmlUrl":35,"ownerBlogUrl":6,"ownerDescription":36,"updatedAt":37,"install":38,"resources":50},"01M79AP6AA8AXHD8V1FZKYARED","tt",null,"See which docs your AI actually discovers with local, zero-token telemetry. Subcommands; \u002Ftt status, \u002Ftt watch [demo|replay], \u002Ftt insights, \u002Ftt instructions, \u002Ftt suggestions, \u002Ftt badge, \u002Ftt gate, \u002Ftt note \u003Ctext>, \u002Ftt doctor, \u002Ftt setup, \u002Ftt uninstall, \u002Ftt help.","Hedde\u002Ftrigger_tree","Hedde","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F1533552?v=4","01M79AP1WDAN52TXWAKW0FP195","Hedde van der Heide","01M79AP1WZ0J44QPJK2H33R3KS",16,"MIT","https:\u002F\u002Fgithub.com\u002FHedde\u002Ftrigger_tree\u002Fblob\u002Fmain\u002Fskills\u002Ftt\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: tt\ndescription: See which docs your AI actually discovers with local, zero-token telemetry. Subcommands; \u002Ftt status, \u002Ftt watch [demo|replay], \u002Ftt insights, \u002Ftt instructions, \u002Ftt suggestions, \u002Ftt badge, \u002Ftt gate, \u002Ftt note \u003Ctext>, \u002Ftt doctor, \u002Ftt setup, \u002Ftt uninstall, \u002Ftt help.\ndisable-model-invocation: true\nallowed-tools: Bash, Read, Write, Artifact\narguments:\n  - subcommand\n  - option\n---\n\n# \u002Ftt — trigger-tree\n\nExecute the subcommand in `$1` (option\u002Ftext in `$2` and beyond; `$ARGUMENTS` holds everything).\n\n**Output discipline — the most important rule:** work silently. No plan up front, no\nintermediate status lines, no explanation of what you are about to do or just did.\nExecute the steps and show only the end result in the format specified per subcommand\nbelow. Report an error in exactly one line.\n\n**Terminology — maturity model:** a file with zero reads is **untouched**, never \"dead\".\nOnly when the stats JSON reports `maturity: \"mature\"` may untouched files be presented\nas **dead-path candidates**. During `cold-start` and `warming` the measurement is simply\ntoo young to judge.\n\nScripts live in `${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002F`. Project data lives in\n`${CLAUDE_PROJECT_DIR}\u002F.trigger-tree\u002F`.\n\n## `$1` = \"help\" or empty\n\nShow exactly this, nothing above or below it:\n\n> **🌳 trigger-tree** — measures which documentation your agents actually use\n>\n> | Command | Does |\n> |---------|------|\n> | `\u002Ftt status` | Snapshot: current heat, lifetime reads, untouched paths |\n> | `\u002Ftt watch` | Live dashboard; recency colors, heat bars, injected context, privacy settings, and focus\u002Fhot\u002Fcold\u002FA–Z\u002FZ–A sorting |\n> | `\u002Ftt watch demo` | Dashboard with synthetic events (writes nothing) |\n> | `\u002Ftt watch replay` | Dashboard replaying the real history, accelerated |\n> | `\u002Ftt insights` | Analysis report: heat\u002Fcold map, untouched paths, hunting, trend + HTML |\n> | `\u002Ftt instructions` | Per-directive adherence, observability, and always-loaded cost |\n> | `\u002Ftt suggestions` | Max 5 prioritized, concrete router fixes — apply after confirmation |\n> | `\u002Ftt badge` | Write `.trigger-tree\u002Fbadge.json`; immature data publishes only `measuring…` |\n> | `\u002Ftt note \u003Ctext>` | Annotate the timeline (e.g. \"sharpened UX router\") — shows up in the trend |\n> | `\u002Ftt doctor` | Verify hooks, privacy, statusline, and telemetry with actionable fixes |\n> | `\u002Ftt setup [truncate\\|hash\\|off]` | Wire the project; choose recognizable previews or privacy-first prompt markers |\n> | `\u002Ftt gate` | Deterministic discoverability score; CI-ready gate with committed baseline |\n> | `\u002Ftt uninstall` | Remove statusline wiring; keep local telemetry for explicit deletion |\n> | `\u002Ftt help` | This overview |\n>\n> Telemetry runs automatically via hooks; the statusline shows the live session counter.\n\n## `$1` = \"status\"\n\n1. Silently run `python3 \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Ftt-stats.py\" --client claude` and read the JSON.\n2. Show only this block:\n\n> **🌳 trigger-tree status** _(period: \u003Cobserved_from> → \u003Cobserved_to>, \u003Csessions> sessions)_\n> \u003Creads> reads · \u003Cscans> searches · \u003Cskill_uses> skill uses · \u003Ctouched_current_files>\u002F\u003Cevaluable_files> current files touched\n> **Health:** \u003Chealth.grade> (\u003Chealth.score>\u002F100) — append \"(provisional)\" unless maturity is `mature`\n>\n> **Current heat:** exclude `state != \"current\"`, then sort `files` by `heat` descending and show the top 5 as\n> `path (h\u003Cheat>, \u003Creads>× lifetime)`, comma-separated.\n> **Untouched:** \u003Ccount of untouched> — followed by the maturity suffix:\n> - `cold-start`: \"(measurement just started — nothing can be called dead yet)\"\n> - `warming`: \"(early signal — more sessions needed before judging)\"\n> - `mature`: \"(dead-path candidates — see \u002Ftt insights)\"\n\n## `$1` = \"watch\"\n\n1. Silently run `TT_CLIENT=claude \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Ftt-open.sh\" $2`\n   (`$2` empty, `demo` or `replay`). The explicit client marker is required so the\n   detached dashboard can load Claude-specific maintenance tips.\n2. Answer with exactly the confirmation line the script prints (it reports tmux split \u002F\n   new window \u002F used terminal). If the script fails, report the cause in one line (on\n   first use macOS may ask for Automation permission; mention that if the error points to it).\n\n## `$1` = \"insights\"\n\n1. Silently run `python3 \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Ftt-stats.py\" --client claude` and read the JSON.\n2. If `maturity` is `cold-start`: answer in one line —\n   `🌳 Not enough data yet (\u003Creads> reads, \u003Csessions> sessions) — check back after a few working sessions.` Stop.\n3. Silently run `python3 \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Ftt-report.py\" --client claude` — it writes\n   `.trigger-tree\u002Freport.html` and prints the path.\n4. Publish that file with the Artifact tool (favicon `🌳`, description \"trigger-tree\n   documentation telemetry report\"). No Artifact tool available → use the local path\n   as the link.\n5. Show only the final report, with a hard maximum of 15 lines:\n\n> **🌳 trigger-tree insights** _(period, #sessions, maturity)_\n>\n> **Health** — grade + score with its three drivers, one line (\"(provisional)\" unless mature).\n> **Key figures** — reads, searches, skill uses, touched current files \u002F evaluable files;\n> mention always-loaded files separately so the inventory equation reconciles.\n> **Agent personas** — when `agents.capture` is `on` and `agents.usage` is non-empty:\n> name the most and least used persona with counts. List `agents.never_invoked` as a\n> review prompt, never a removal recommendation, and say nothing about never-invoked when\n> `agents.never_invoked_status` is `awaiting-capture`. When `agents.capture` is `off` and\n> `agents.defined` is above zero, say instead that N personas are defined but usage is not\n> being recorded, and that `\u002Ftt setup` enables it; do not imply they are unused.\n> **Folder heat\u002Fcold map** — from `folders`: name the hottest folder by current\n> decayed `heat` (also state 30-day and lifetime reads) and the least-covered folder,\n> one line each. Cold means inactive now, not obsolete.\n> **Untouched paths** — when `mature`: one line per path with a category — 🗑 remove\u002Fmerge,\n> 🧭 sharpen router (with a concrete proposal), 📦 intentional archive. Use\n> `untouched_detail`: a path with zero `inbound_refs` is a **router gap** (no doc\n> links to it) — that's nearly always 🧭. `is_router: true` is an unread router signal,\n> never a template. `template: true` may support 📦 but is not a removal verdict.\n> When `warming`: present as untouched with the note that judgment needs more data;\n> no categories.\n> **Trend** — only when `trend` has 2+ periods: is `search_ratio` falling or rising,\n> and does that correlate with any `notes` (router changes)? Never say a note or edit\n> caused the change.\n> **Search activity** — only if scans > 20% of reads: use `search_activity` to report\n> the tool mix, sessions\u002Ftotal_sessions, and `pattern`. A search event is an explicit search in a\n> watched path, not proof of failed routing. Treat `concentrated` activity as possible\n> bulk\u002Fresearch workflow; only `distributed` recurrence may support a routing hypothesis.\n> **Task clusters** — top 2-3 from `clusters`: \"tasks like \u003Cexample prompt> consistently\n> use \u003Cpaths>\" — flag when a cluster misses an obvious doc.\n> **Router proposals** — zero to three, never a quota. Use only `router_coverage`\n> entries where both `router` and target already exist and the target is in `unlisted`.\n> Before presenting a proposal, silently verify against the repository that the router\n> still exists and still does not mention the target. If verification fails or evidence\n> is insufficient, omit it. “No evidence-backed router proposals” is a valid result.\n>\n> 📊 Full report: \u003Cartifact link or file path>\n\nAnalysis rules (do not repeat them in the output): lifetime read counts never decay;\ncurrent `heat` uses the `heat_model` 30-day half-life and is distinct from untouched;\nread counts and heat are signals, not verdicts;\nthe schema's `scan` counts are narrow search telemetry and do not establish why a search occurred;\nfiles in `always_loaded_inventory` are never dead by definition (system-prompt injection);\n`retired_files` are historical evidence and must not appear in current rankings; a file\nyounger than the measurement period is new, not untouched; subagent reads (the `agents`\nfield) count fully; skill uses make `.claude\u002Fskills\u002F**` measurable — an invoked skill's\nSKILL.md counts as touched.\n\n## `$1` = \"suggestions\"\n\nSilently run `python3 \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Ftt-suggestions.py\" --no-apply-prompt`\n(fall back to `python` if needed). The script keeps full stats off stdout and changes\nnothing. Its output is tiered: at most five numbered, appliable router edits (each\nverified against existing files, with evidence numbers), up to two unnumbered\n\"Worth a look\" observations, and one summary line for low-read but likely-critical\nfiles — and the tool result already shows it to the user, so **do not repeat or\nreformat the script output**. Reply with exactly one line: `Apply any of these?\nReply with the numbers.` when numbered edits exist, otherwise `Nothing to apply.`\nApply only numbered edits the user explicitly confirms, then suggest recording the\nchange with `\u002Ftt note`.\n\n## `$1` = \"instructions\"\n\nSilently run `python3 \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Ftt-instructions.py\"` (fall back\nto `python` if needed), passing through every argument after `instructions`, including\n`--init`, `--selftest`, `--explain ID`, `--json`, `--check`, `--min-rate N`, and `--min-measured N`. Show the script\noutput verbatim; do not summarize or reformat it. `--init` creates or refreshes only a\nhash-bound scaffold. Never invent, add, or apply probes without showing the proposed\ndirective entries and receiving explicit user confirmation. Treat `unobserved` as missing\nevidence, never as a violation; capture-disabled probes are excluded from rates.\n\n## `$1` = \"tips\"\n\nSilently run `python3 \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Ftt-tips.py\" --client claude`\n(fall back to `python` if needed) and show its concise output verbatim. These are\nmaintenance recommendations only; do not edit memory or instruction files automatically.\n\n## `$1` = \"badge\"\n\nSilently run `python3 \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Ftt-stats.py\" --client claude --badge`\nand answer with exactly one line: `🌳 Docs-health badge written to \u003Cpath>.` The endpoint\nJSON never exposes a provisional letter; before mature measurement it says `measuring…`.\n\n## `$1` = \"note\"\n\n1. Take everything after the subcommand as the note text. Silently run:\n   `python3 \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Ftt-log.py\" note \"\u003Ctext>\"`\n   (the session id is picked up automatically from CLAUDE_CODE_SESSION_ID, or CLAUDE_SESSION_ID\n   \u002F TT_SESSION_ID when a client or hook sets those instead).\n2. Answer with exactly one line: `🌳 Noted: \"\u003Ctext>\" — it will show up in the trend timeline.`\n   Empty text → `Usage: \u002Ftt note \u003Ctext>` in one line.\n\n## `$1` = \"setup\"\n\n1. If `$2` is absent, ask which local prompt mode to use: `truncate` (recommended;\n   first 200 characters), `hash` (no prompt text), or `off` (marker only). Then run\n   `python3 \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Ftt-setup.py\" --prompt-mode \"$MODE\"`.\n   If `$2` is already `truncate`, `hash`, or `off`, use it without asking. An existing\n   config is preserved only when the script is run without an explicit mode.\n2. Show the script's summary lines verbatim (they report created\u002Fupdated\u002Fskipped per\n   step), nothing else.\n\n## `$1` = \"doctor\"\n\nSilently run `python3 \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Ftt-doctor.py\"` (fall back to\n`python` if needed). Show its output verbatim. A failed check is a diagnostic result,\nnot a skill error: still show the complete output.\n\n## `$1` = \"gate\"\n\nSilently run `python3 \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Ftt-gate.py\"` (fall back to\n`python` if needed), passing through any extra arguments (`--update-baseline`,\n`--min-score`, `--badge`). The score, components, and findings are static repository\nfacts — no telemetry — and the tool result already shows them; do not repeat them.\nReply with exactly one line: `Gate passed — score \u003CN>%.` or\n`Gate failed — see the findings above.`\n\n## `$1` = \"uninstall\"\n\nSilently run `python3 \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Ftt-uninstall.py\"` and show its\nsummary verbatim. It removes only trigger-tree's statusline registration and copied\nscript; `.trigger-tree\u002F` telemetry and `.gitignore` entries remain for explicit deletion.\n\n## Any other `$1`\n\nAnswer with exactly one line: `Unknown subcommand '$1' — try \u002Ftt help.`\n","# \u002Ftt — trigger-tree\n\nExecute the subcommand in `$1` (option\u002Ftext in `$2` and beyond; `$ARGUMENTS` holds everything).\n\n**Output discipline — the most important rule:** work silently. No plan up front, no\nintermediate status lines, no explanation of what you are about to do or just did.\nExecute the steps and show only the end result in the format specified per subcommand\nbelow. Report an error in exactly one line.\n\n**Terminology — maturity model:** a file with zero reads is **untouched**, never \"dead\".\nOnly when the stats JSON reports `maturity: \"mature\"` may untouched files be presented\nas **dead-path candidates**. During `cold-start` and `warming` the measurement is simply\ntoo young to judge.\n\nScripts live in `${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002F`. Project data lives in\n`${CLAUDE_PROJECT_DIR}\u002F.trigger-tree\u002F`.\n\n## `$1` = \"help\" or empty\n\nShow exactly this, nothing above or below it:\n\n> **🌳 trigger-tree** — measures which documentation your agents actually use\n>\n> | Command | Does |\n> |---------|------|\n> | `\u002Ftt status` | Snapshot: current heat, lifetime reads, untouched paths |\n> | `\u002Ftt watch` | Live dashboard; recency colors, heat bars, injected context, privacy settings, and focus\u002Fhot\u002Fcold\u002FA–Z\u002FZ–A sorting |\n> | `\u002Ftt watch demo` | Dashboard with synthetic events (writes nothing) |\n> | `\u002Ftt watch replay` | Dashboard replaying the real history, accelerated |\n> | `\u002Ftt insights` | Analysis report: heat\u002Fcold map, untouched paths, hunting, trend + HTML |\n> | `\u002Ftt instructions` | Per-directive adherence, observability, and always-loaded cost |\n> | `\u002Ftt suggestions` | Max 5 prioritized, concrete router fixes — apply after confirmation |\n> | `\u002Ftt badge` | Write `.trigger-tree\u002Fbadge.json`; immature data publishes only `measuring…` |\n> | `\u002Ftt note \u003Ctext>` | Annotate the timeline (e.g. \"sharpened UX router\") — shows up in the trend |\n> | `\u002Ftt doctor` | Verify hooks, privacy, statusline, and telemetry with actionable fixes |\n> | `\u002Ftt setup [truncate\\|hash\\|off]` | Wire the project; choose recognizable previews or privacy-first prompt markers |\n> | `\u002Ftt gate` | Deterministic discoverability score; CI-ready gate with committed baseline |\n> | `\u002Ftt uninstall` | Remove statusline wiring; keep local telemetry for explicit deletion |\n> | `\u002Ftt help` | This overview |\n>\n> Telemetry runs automatically via hooks; the statusline shows the live session counter.\n\n## `$1` = \"status\"\n\n1. Silently run `python3 \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Ftt-stats.py\" --client claude` and read the JSON.\n2. Show only this block:\n\n> **🌳 trigger-tree status** _(period: \u003Cobserved_from> → \u003Cobserved_to>, \u003Csessions> sessions)_\n> \u003Creads> reads · \u003Cscans> searches · \u003Cskill_uses> skill uses · \u003Ctouched_current_files>\u002F\u003Cevaluable_files> current files touched\n> **Health:** \u003Chealth.grade> (\u003Chealth.score>\u002F100) — append \"(provisional)\" unless maturity is `mature`\n>\n> **Current heat:** exclude `state != \"current\"`, then sort `files` by `heat` descending and show the top 5 as\n> `path (h\u003Cheat>, \u003Creads>× lifetime)`, comma-separated.\n> **Untouched:** \u003Ccount of untouched> — followed by the maturity suffix:\n> - `cold-start`: \"(measurement just started — nothing can be called dead yet)\"\n> - `warming`: \"(early signal — more sessions needed before judging)\"\n> - `mature`: \"(dead-path candidates — see \u002Ftt insights)\"\n\n## `$1` = \"watch\"\n\n1. Silently run `TT_CLIENT=claude \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Ftt-open.sh\" $2`\n   (`$2` empty, `demo` or `replay`). The explicit client marker is required so the\n   detached dashboard can load Claude-specific maintenance tips.\n2. Answer with exactly the confirmation line the script prints (it reports tmux split \u002F\n   new window \u002F used terminal). If the script fails, report the cause in one line (on\n   first use macOS may ask for Automation permission; mention that if the error points to it).\n\n## `$1` = \"insights\"\n\n1. Silently run `python3 \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Ftt-stats.py\" --client claude` and read the JSON.\n2. If `maturity` is `cold-start`: answer in one line —\n   `🌳 Not enough data yet (\u003Creads> reads, \u003Csessions> sessions) — check back after a few working sessions.` Stop.\n3. Silently run `python3 \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Ftt-report.py\" --client claude` — it writes\n   `.trigger-tree\u002Freport.html` and prints the path.\n4. Publish that file with the Artifact tool (favicon `🌳`, description \"trigger-tree\n   documentation telemetry report\"). No Artifact tool available → use the local path\n   as the link.\n5. Show only the final report, with a hard maximum of 15 lines:\n\n> **🌳 trigger-tree insights** _(period, #sessions, maturity)_\n>\n> **Health** — grade + score with its three drivers, one line (\"(provisional)\" unless mature).\n> **Key figures** — reads, searches, skill uses, touched current files \u002F evaluable files;\n> mention always-loaded files separately so the inventory equation reconciles.\n> **Agent personas** — when `agents.capture` is `on` and `agents.usage` is non-empty:\n> name the most and least used persona with counts. List `agents.never_invoked` as a\n> review prompt, never a removal recommendation, and say nothing about never-invoked when\n> `agents.never_invoked_status` is `awaiting-capture`. When `agents.capture` is `off` and\n> `agents.defined` is above zero, say instead that N personas are defined but usage is not\n> being recorded, and that `\u002Ftt setup` enables it; do not imply they are unused.\n> **Folder heat\u002Fcold map** — from `folders`: name the hottest folder by current\n> decayed `heat` (also state 30-day and lifetime reads) and the least-covered folder,\n> one line each. Cold means inactive now, not obsolete.\n> **Untouched paths** — when `mature`: one line per path with a category — 🗑 remove\u002Fmerge,\n> 🧭 sharpen router (with a concrete proposal), 📦 intentional archive. Use\n> `untouched_detail`: a path with zero `inbound_refs` is a **router gap** (no doc\n> links to it) — that's nearly always 🧭. `is_router: true` is an unread router signal,\n> never a template. `template: true` may support 📦 but is not a removal verdict.\n> When `warming`: present as untouched with the note that judgment needs more data;\n> no categories.\n> **Trend** — only when `trend` has 2+ periods: is `search_ratio` falling or rising,\n> and does that correlate with any `notes` (router changes)? Never say a note or edit\n> caused the change.\n> **Search activity** — only if scans > 20% of reads: use `search_activity` to report\n> the tool mix, sessions\u002Ftotal_sessions, and `pattern`. A search event is an explicit search in a\n> watched path, not proof of failed routing. Treat `concentrated` activity as possible\n> bulk\u002Fresearch workflow; only `distributed` recurrence may support a routing hypothesis.\n> **Task clusters** — top 2-3 from `clusters`: \"tasks like \u003Cexample prompt> consistently\n> use \u003Cpaths>\" — flag when a cluster misses an obvious doc.\n> **Router proposals** — zero to three, never a quota. Use only `router_coverage`\n> entries where both `router` and target already exist and the target is in `unlisted`.\n> Before presenting a proposal, silently verify against the repository that the router\n> still exists and still does not mention the target. If verification fails or evidence\n> is insufficient, omit it. “No evidence-backed router proposals” is a valid result.\n>\n> 📊 Full report: \u003Cartifact link or file path>\n\nAnalysis rules (do not repeat them in the output): lifetime read counts never decay;\ncurrent `heat` uses the `heat_model` 30-day half-life and is distinct from untouched;\nread counts and heat are signals, not verdicts;\nthe schema's `scan` counts are narrow search telemetry and do not establish why a search occurred;\nfiles in `always_loaded_inventory` are never dead by definition (system-prompt injection);\n`retired_files` are historical evidence and must not appear in current rankings; a file\nyounger than the measurement period is new, not untouched; subagent reads (the `agents`\nfield) count fully; skill uses make `.claude\u002Fskills\u002F**` measurable — an invoked skill's\nSKILL.md counts as touched.\n\n## `$1` = \"suggestions\"\n\nSilently run `python3 \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Ftt-suggestions.py\" --no-apply-prompt`\n(fall back to `python` if needed). The script keeps full stats off stdout and changes\nnothing. Its output is tiered: at most five numbered, appliable router edits (each\nverified against existing files, with evidence numbers), up to two unnumbered\n\"Worth a look\" observations, and one summary line for low-read but likely-critical\nfiles — and the tool result already shows it to the user, so **do not repeat or\nreformat the script output**. Reply with exactly one line: `Apply any of these?\nReply with the numbers.` when numbered edits exist, otherwise `Nothing to apply.`\nApply only numbered edits the user explicitly confirms, then suggest recording the\nchange with `\u002Ftt note`.\n\n## `$1` = \"instructions\"\n\nSilently run `python3 \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Ftt-instructions.py\"` (fall back\nto `python` if needed), passing through every argument after `instructions`, including\n`--init`, `--selftest`, `--explain ID`, `--json`, `--check`, `--min-rate N`, and `--min-measured N`. Show the script\noutput verbatim; do not summarize or reformat it. `--init` creates or refreshes only a\nhash-bound scaffold. Never invent, add, or apply probes without showing the proposed\ndirective entries and receiving explicit user confirmation. Treat `unobserved` as missing\nevidence, never as a violation; capture-disabled probes are excluded from rates.\n\n## `$1` = \"tips\"\n\nSilently run `python3 \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Ftt-tips.py\" --client claude`\n(fall back to `python` if needed) and show its concise output verbatim. These are\nmaintenance recommendations only; do not edit memory or instruction files automatically.\n\n## `$1` = \"badge\"\n\nSilently run `python3 \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Ftt-stats.py\" --client claude --badge`\nand answer with exactly one line: `🌳 Docs-health badge written to \u003Cpath>.` The endpoint\nJSON never exposes a provisional letter; before mature measurement it says `measuring…`.\n\n## `$1` = \"note\"\n\n1. Take everything after the subcommand as the note text. Silently run:\n   `python3 \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Ftt-log.py\" note \"\u003Ctext>\"`\n   (the session id is picked up automatically from CLAUDE_CODE_SESSION_ID, or CLAUDE_SESSION_ID\n   \u002F TT_SESSION_ID when a client or hook sets those instead).\n2. Answer with exactly one line: `🌳 Noted: \"\u003Ctext>\" — it will show up in the trend timeline.`\n   Empty text → `Usage: \u002Ftt note \u003Ctext>` in one line.\n\n## `$1` = \"setup\"\n\n1. If `$2` is absent, ask which local prompt mode to use: `truncate` (recommended;\n   first 200 characters), `hash` (no prompt text), or `off` (marker only). Then run\n   `python3 \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Ftt-setup.py\" --prompt-mode \"$MODE\"`.\n   If `$2` is already `truncate`, `hash`, or `off`, use it without asking. An existing\n   config is preserved only when the script is run without an explicit mode.\n2. Show the script's summary lines verbatim (they report created\u002Fupdated\u002Fskipped per\n   step), nothing else.\n\n## `$1` = \"doctor\"\n\nSilently run `python3 \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Ftt-doctor.py\"` (fall back to\n`python` if needed). Show its output verbatim. A failed check is a diagnostic result,\nnot a skill error: still show the complete output.\n\n## `$1` = \"gate\"\n\nSilently run `python3 \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Ftt-gate.py\"` (fall back to\n`python` if needed), passing through any extra arguments (`--update-baseline`,\n`--min-score`, `--badge`). The score, components, and findings are static repository\nfacts — no telemetry — and the tool result already shows them; do not repeat them.\nReply with exactly one line: `Gate passed — score \u003CN>%.` or\n`Gate failed — see the findings above.`\n\n## `$1` = \"uninstall\"\n\nSilently run `python3 \"${CLAUDE_PLUGIN_ROOT}\u002Fscripts\u002Ftt-uninstall.py\"` and show its\nsummary verbatim. It removes only trigger-tree's statusline registration and copied\nscript; `.trigger-tree\u002F` telemetry and `.gitignore` entries remain for explicit deletion.\n\n## Any other `$1`\n\nAnswer with exactly one line: `Unknown subcommand '$1' — try \u002Ftt help.`\n",{"name":5,"arguments":24,"description":7,"allowed-tools":27,"disable-model-invocation":28},[25,26],"subcommand","option","Bash, Read, Write, Artifact",true,"main","9f2eb03fcf264a22fc80e279f21073c42d49b92b","skills\u002Ftt",1,"https:\u002F\u002Fgithub.com\u002FHedde\u002Ftrigger_tree","https:\u002F\u002Fhedde.github.io\u002Ftrigger_tree\u002F","https:\u002F\u002Fgithub.com\u002FHedde","programmer, hacker and sceptic. more knowledge, less understanding.","2026-11-02 16:13:23",{"source":8,"skillName":5,"commands":39,"prompt":49,"zipAvailable":28},[40,43,46],{"tool":41,"command":42},"npx","npx skills add Hedde\u002Ftrigger_tree --skill tt",{"tool":44,"command":45},"bunx","bunx skills add Hedde\u002Ftrigger_tree --skill tt",{"tool":47,"command":48},"pnpm","pnpm dlx skills add Hedde\u002Ftrigger_tree --skill tt","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FHedde\u002Ftrigger_tree\u002Fblob\u002Fmain\u002Fskills\u002Ftt\u002FSKILL.md",[],[52,58],{"skillNo":53,"name":54,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":55,"categories":56,"keywords":57,"publishedAt":19},"01M79APAN5SK133Q04ETWZR3JK","deploy","https:\u002F\u002Fgithub.com\u002FHedde\u002Ftrigger_tree\u002Fblob\u002Fmain\u002Ftests\u002Ffixture-project\u002F.claude\u002Fskills\u002Fdeploy\u002FSKILL.md",[],[],{"skillNo":59,"name":60,"nameZh":6,"description":61,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":62,"categories":63,"keywords":64,"publishedAt":19},"01M79AP3RCNHZRHZC8PTNAZ52A","trigger-tree","Inspect and improve documentation discovery using trigger-tree's local telemetry. Use for trigger-tree status, live dashboard, heat\u002Fcold maps, documentation health, router suggestions, setup, or diagnostics.","https:\u002F\u002Fgithub.com\u002FHedde\u002Ftrigger_tree\u002Fblob\u002Fmain\u002Fcodex-skills\u002Ftrigger-tree\u002FSKILL.md",[],[]]