Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/thkt/dotclaude/ablatenpx skills add thkt/dotclaude --skill ablategit clone --depth 1 https://github.com/thkt/dotclaudeWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00057 | $0.01236 |
| Opus 5 | $0.00028 | $0.00618 |
| Sonnet 5 | $0.00011 | $0.00247 |
| Haiku 4.5 | $0.00006 | $0.00124 |
Grade A, and why
ablate scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 56 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/ablate - ハーネスの片側アブレーション
Input
$ARGUMENTS は測定対象を 1 つに絞る要素のパス。省略したときは Phase 1 が列挙した全要素を対象にする。
判定と閾値の所在
アームの一覧、1 アームあたりの実行回数、通過閾値はすべて ${CLAUDE_SKILL_DIR}/scripts/arms.py の定数が持つ。分類の基準は ${CLAUDE_SKILL_DIR}/scripts/verdict.py が持つ。DR ゲートの基準、未達の記録を表す印、記録を読む先は ${CLAUDE_SKILL_DIR}/scripts/dr_gate.py が持つ。計測窓と rare-by-design の集合は ${CLAUDE_SKILL_DIR}/scripts/usage_counts.py が持つ。規則ごとの起動タスクと固定タスクセットは ${CLAUDE_SKILL_DIR}/references/measurement-criteria.md が持つ。この本文に数値を書き写さない (docs/wiki/deterministic-script-judgment.md)。
Phase 1: 列挙
skills/_lib/harness_elements.py の enumerate_elements(root) を呼び、ハーネス要素とその分類を得る。$ARGUMENTS が要素のパスを指すときは、その 1 件だけを Phase 2 へ渡す。
python3 -c 'import sys; sys.path.insert(0, "skills/_lib"); import harness_elements, json; print(json.dumps(harness_elements.enumerate_elements(".")))'
Phase 2: アーム実行
Phase 1 が返した要素それぞれについて、arms.ARMS の各アームで arms.arm_command(arm, element) が返す命令を組み、arms.RUN_COUNT 回実行する。その要素の起動タスクは ${CLAUDE_SKILL_DIR}/references/measurement-criteria.md から取る。各 run の結果から、その要素についての観測 1 件を組む。
| 状況 | 扱い |
|---|---|
run 数が arms.RUN_COUNT に届かない |
arms.measurement_status(runs) が unmeasured を返すまま進める |
wiped+1 に渡す要素が決まらない |
arm_command が ValueError で止まるので、要素を確定してから呼ぶ |
| 実行が失敗し結果を読めない run がある | その run を数えず、observation に届いた run 数だけを載せる |
Phase 3: レポート
report.write_report(root, observations) を呼ぶ。削除候補がレポートへ届く前に dr_gate.gate が docs/decisions/ を読み、生きている記録が支配する要素を保留するので、Summary はその件数を分けて数える。同時に usage_counts.py も実行し、各要素の発火回数と最終使用日を同じ Harness Elements 表へ組み込む。見る経路は 1 つで、別経路を並走させない。書き出し先の既定は docs/audit/ で、ファイル名は UTC の <YYYY-MM-DD>-<HHMMSS>-ablate.md。節の順は ${CLAUDE_SKILL_DIR}/templates/report-template.md が持つ。
python3 -c 'import sys; sys.path.insert(0, "skills/ablate/scripts"); sys.path.insert(0, "skills/_lib"); import report, json, pathlib; print(report.write_report(pathlib.Path("."), json.load(sys.stdin)))' < <observations.json>
What ships with it
8 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 56 lines · 57 tokens per session scan A 80c56b17feb3
ablate is a skill published in the GitHub repository thkt/dotclaude (11 stars, last pushed 2d ago), licensed MIT. It adds 57 tokens to every session and 1,236 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
create-modal
Create declarative modals using the modal library API. Covers modal types (confirm, input, select, form), sections (Text, Buttons, Input, Textarea, Checkbox, List, Combo, When, Custom), rendering with OverlayModal, and keyboard/mouse handling. Use when adding modals or dialogs to the application.
create-plugin
Create new sidecar plugins implementing the plugin.Plugin interface, rendering views with Bubble Tea, handling keyboard input via keymap contexts, and integrating with the app shell (footer hints, event bus, adapters). Use when creating a new plugin, modifying plugin architecture, or debugging plugin…
merge-strategy
Git merge strategies, conflict resolution approaches, merge vs rebase recommendations, and branch integration patterns in sidecar. Covers pull strategy menu, direct merge workflow, squash merge, commit message templates, configurable defaults, and protected branches. Use when working on git merge features or making…
autoprompt
Explicit-only useful-first orchestration. Invoke /autoprompt to turn a mission into one executable roadmap, build dependency-safe lanes, and verify the result with independent reviewers. Never infer invocation from ordinary requests. Never resume from leftover artifacts without an explicit resume instruction.
autoprompt
Explicit-only useful-first orchestration. Invoke /autoprompt to turn a mission into one executable roadmap, build dependency-safe lanes, and verify the result with independent reviewers. Never infer invocation from ordinary requests. Never resume from leftover artifacts without an explicit resume instruction.
autoprompt
Explicit-only useful-first orchestration. Invoke /autoprompt to turn a mission into one executable roadmap, build dependency-safe lanes, and verify the result with independent reviewers. Never infer invocation from ordinary requests. Never resume from leftover artifacts without an explicit resume instruction.