[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M75G47BYNP52TAFC2MCB8TJE":3,"skill-same:01M75G47BYNP52TAFC2MCB8TJE":56},{"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":33,"commitSha":34,"skillDirectory":35,"forksCount":36,"repositoryHtmlUrl":37,"repositoryHomepageUrl":6,"ownerHtmlUrl":38,"ownerBlogUrl":39,"ownerDescription":40,"updatedAt":41,"install":42,"resources":55},"01M75G47BYNP52TAFC2MCB8TJE","agentbench",null,"Benchmark your OpenClaw agent across 40 real-world tasks. Tests file creation, research, data analysis, multi-step workflows, memory, error handling, and tool efficiency. Not a coding benchmark — measures your agent setup and config.","javimosch\u002Fopen-claw-skills","javimosch","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F12900508?v=4","01M6D3F2AS6PV75ET9PZ1QGP8J","Javier Leandro Arancibia","01M75B8NCTAJG9WH5YNZH7C3D8",0,"MIT","https:\u002F\u002Fgithub.com\u002Fjavimosch\u002Fopen-claw-skills\u002Fblob\u002Fmain\u002Fskills\u002Fexe215\u002Fagentbench\u002FSKILL.md",[],[],"2026-11-12 13:55:43",false,"---\nname: agentbench\ndescription: Benchmark your OpenClaw agent across 40 real-world tasks. Tests file creation, research, data analysis, multi-step workflows, memory, error handling, and tool efficiency. Not a coding benchmark — measures your agent setup and config.\nhomepage: https:\u002F\u002Fwww.agentbench.app\nmetadata: { \"openclaw\": { \"emoji\": \"📊\", \"requires\": { \"bins\": [\"jq\", \"bash\", \"python3\"] } } }\n---\n\n# AgentBench for OpenClaw\n\nBenchmark your OpenClaw agent's general capabilities across 40 real-world tasks spanning 7 domains.\n\n## Commands\n\nWhen the user says any of these, follow the corresponding instructions:\n\n- **`\u002Fbenchmark`** — Run the full benchmark suite (all 40 tasks)\n- **`\u002Fbenchmark --fast`** — Run only easy+medium tasks (19 tasks)\n- **`\u002Fbenchmark --suite \u003Cname>`** — Run one domain only\n- **`\u002Fbenchmark --task \u003Cid>`** — Run a single task\n- **`\u002Fbenchmark --strict`** — Tag results as externally verified scoring\n- **`\u002Fbenchmark-list`** — List all tasks grouped by domain\n- **`\u002Fbenchmark-results`** — Show results from previous runs\n- **`\u002Fbenchmark-compare`** — Compare two runs side-by-side\n\nFlags are combinable: `\u002Fbenchmark --fast --suite research`\n\n## Running a Benchmark\n\n### Step 1: Discover Tasks\n\nRead task.yaml files from the `tasks\u002F` directory in this skill:\n\n```\ntasks\u002F{suite-name}\u002F{task-name}\u002Ftask.yaml\n```\n\nEach task.yaml contains: name, id, suite, difficulty, mode, user_message, input_files, expected_outputs, expected_metrics, scoring weights.\n\nFilter by `--suite` or `--task` if specified. If `--fast` is set and `--task` is not, filter to only tasks where difficulty is \"easy\" or \"medium\".\n\nProfile is \"fast\" if `--fast` was specified, otherwise \"full\".\n\nList discovered tasks with count and suites.\n\n### Step 2: Set Up Run Directory\n\nGenerate a run ID from the current timestamp: `YYYYMMDD-HHmmss`\n\nRead `suite_version` from `skill.json` in this skill directory.\n\nCreate the results directory:\n```\nagentbench-results\u002F{run-id}\u002F\n```\n\nAnnounce: `Starting AgentBench run {run-id} | Profile: {profile} | Suite version: {suite_version} | Tasks: {count}`\n\n### Step 3: Execute Each Task\n\nFor each task:\n\n1. **Set up workspace**:\n   - Create `\u002Ftmp\u002Fagentbench-task-{task-id}\u002F` as workspace\n   - Copy input files from `tasks\u002F{suite}\u002F{task}\u002Finputs\u002F` to the workspace (if inputs\u002F exists)\n   - If the task directory contains a `setup.sh`: run `bash tasks\u002F{suite}\u002F{task}\u002Fsetup.sh {workspace-path}`\n   - For `file-unchanged` validators: compute checksums of specified files after setup, before task execution\n\n2. **Announce**: `Running: {task.name} [{task.suite}] (difficulty: {task.difficulty})`\n\n3. **Record start time** (milliseconds): `date +%s%3N`\n\n4. **Execute the task yourself directly**:\n   - Read the task's `user_message` and execute it as if a real user sent you the request\n   - Work ONLY within the workspace directory\n   - If input files are listed, read them from the workspace\n   - Execute naturally — use the appropriate tools (read, write, edit, exec, web_search, web_fetch, etc.)\n   - Create any output files in the workspace directory\n   - When done, write a brief `execution-trace.md` to the workspace:\n     - What you understood the task to be\n     - What approach you took\n     - What files you created or modified\n     - Any difficulties or decisions you made\n\n5. **Record end time** and compute duration\n\n6. **Collect metrics**:\n   - `total_time_ms`: end - start\n   - `tool_calls_total`: count how many tool calls you made during this task\n   - `errors`: count any tool call failures\n   - `planning_ratio`: estimate the fraction of time spent reading\u002Fthinking vs producing output (approximate is fine)\n\n7. **Layer 0 — Automated Structural Checks** (compute directly):\n   After task execution, check the workspace. For each entry in `expected_outputs`:\n   - `file-exists`: Check if file exists. 30 points if found, 0 if not.\n   - `content-contains`: Read file, check each required section keyword (case-insensitive). Points proportional to matches found. Pool: 40 points.\n   - `word-count-range`: Count words. In range = 30 points. Within 2x range = 15 points. Outside = 0.\n   - `git-log-contains`: Check git log for expected strings. 30 points if all found, proportional otherwise.\n   - `directory-structure`: Check all paths exist. 30 points if all present, proportional for partial.\n   - `command-output-contains`: Run command, check output contains all strings. 30 points if match, 0 if not.\n   - `file-unchanged`: Compare checksum against pre-execution checksum. 30 points if unchanged, 0 if modified.\n   - `link-consistency`: Scan files for link syntax consistency. 30 points if consistent, 15 if mostly consistent (>70% one style), 0 if mixed.\n   - Normalize total to 0-100.\n\n8. **Layer 1 — Metrics Analysis** (compute directly):\n   If task has expected_metrics:\n   - Tool calls within expected range: 40 points\n   - Tool calls within 2x range: 20 points\n   - Outside 2x range: 0 points\n   - Planning ratio within expected range: 30 points\n   - Planning ratio outside but within 2x: 15 points\n   - Way off: 0 points\n   - Zero errors: 30 points\n   - 1-2 errors: 15 points\n   - 3+ errors: 0 points\n   - Normalize to 0-100. If no metrics available, score as 50.\n   - Token estimate is tracked for reporting but NOT scored.\n\n9. **Layer 2 — Behavioral Analysis** (self-evaluate honestly, 0-100):\n   Score based on HOW you executed:\n\n   **Instruction Adherence (30 points):**\n   - 30: Followed all instructions precisely\n   - 20: Mostly followed, minor deviations\n   - 10: Significant deviations\n   - 0: Ignored or misunderstood\n\n   **Tool Appropriateness (25 points)** — rule-based first:\n   - Penalty: -10 for each use of `exec cat` instead of `read` to read files\n   - Penalty: -10 for each use of `exec echo\u002Fprintf` instead of `write` to create files\n   - Penalty: -5 for each use of `exec sed\u002Fawk` instead of `edit` for file edits\n   - Start at 25, apply penalties, floor at 0\n\n   **Approach Quality (25 points)** — check read-before-write:\n   - 25: Read all inputs before producing output\n   - 15: Read most inputs, minor gaps\n   - 5: Started producing output without reading context\n   - 0: No clear approach\n\n   **Error Recovery (20 points):**\n   - 20: Clean recovery or no errors occurred\n   - 10: Partial recovery\n   - 0: Failed to recover\n\n10. **Layer 3 — Output Quality** (self-evaluate honestly, 0-100):\n    Score the deliverable:\n\n    **Completeness (25):** All requirements met? Gaps?\n    **Accuracy (25):** Content correct? Calculations right?\n    **Formatting (25):** Well-structured? Correct file format?\n    **Polish (25):** Would a user be satisfied?\n\n11. **Compute composite score**:\n    ```\n    score = (L0 × 0.20) + (L1 × 0.35) + (L2 × 0.20) + (L3 × 0.25)\n    ```\n    Use weights from task.yaml if specified, otherwise these defaults.\n\n12. **Save task result** to `agentbench-results\u002F{run-id}\u002F{task-id}\u002F`:\n    - `scores.json`: All layer scores, composite, breakdown, notes\n    - `metrics.json`: Timing, tool calls, errors, planning ratio\n    - Copy output files\n\n13. **Display**: `{task.name}: {composite}\u002F100 (L0:{l0} L1:{l1} L2:{l2} L3:{l3})`\n\n### Step 4: Generate Report\n\nAfter all tasks:\n\n1. Compute domain averages (group by suite, average composite scores)\n2. Compute overall score (average of domain scores — equal domain weighting)\n3. Compute aggregate metrics\n\nGenerate three files in `agentbench-results\u002F{run-id}\u002F`:\n\n**results.json** — Machine-readable with this structure:\n```json\n{\n  \"run_id\": \"20260222-143022\",\n  \"timestamp\": \"2026-02-22T14:30:22Z\",\n  \"platform\": \"openclaw\",\n  \"mode\": \"sandboxed\",\n  \"profile\": \"full\",\n  \"suite_version\": \"1.0.0\",\n  \"scoring_method\": \"self-scored\",\n  \"overall_score\": 74,\n  \"duration_ms\": 754000,\n  \"task_count\": 40,\n  \"metrics\": {\n    \"total_tool_calls\": 187,\n    \"total_errors\": 3,\n    \"avg_planning_ratio\": 0.28,\n    \"est_tokens\": 245000\n  },\n  \"domain_scores\": {},\n  \"tasks\": []\n}\n```\n\nIf `--strict` was used, set `scoring_method` to `\"externally-verified\"`.\n\n**Integrity signature**: After building results.json (without signature field), compute:\n```bash\nSIG=$(echo -n \"$CONTENT\" | openssl dgst -sha256 -hmac \"agentbench-v1-{run_id}-{suite_version}-integrity\" | awk '{print $2}')\n```\nAdd as `\"signature\"` field to results.json.\n\n**report.md** — Markdown summary: Overall Score, Metrics, Domain Breakdown, Task Details, Top Failures, Recommendations.\n\n**report.html** — Self-contained HTML dashboard (inline CSS\u002FJS, no external deps):\n- Score display with color (green 80+, yellow 60-79, red \u003C60)\n- Domain cards with score bars\n- Task detail table (sortable, expandable)\n- Top failures section\n- Dark mode via prefers-color-scheme\n- Footer: \"Generated by AgentBench v1.0.0 (OpenClaw) | Suite v{suite_version} | Profile: {profile}\"\n\n### Step 5: Present Results\n\n1. Display overall score\n2. Show domain breakdown\n3. Tell user where results are saved\n4. Mention they can submit to https:\u002F\u002Fwww.agentbench.app\u002Fsubmit\n\n### Step 6: Clean Up\n\nRun teardown.sh if present. Remove temp workspace directories unless `--keep-workspace` was specified.\n\n## Listing Tasks (`\u002Fbenchmark-list`)\n\nRead all task.yaml files, group by suite, display as:\n```\n## file-creation (9 tasks)\n  - project-scaffold [easy]\n  - project-proposal [medium]\n  ...\n```\n\n## Viewing Results (`\u002Fbenchmark-results`)\n\nList all directories in `agentbench-results\u002F`, show run ID, date, overall score, profile, and task count for each.\n\n## Comparing Runs (`\u002Fbenchmark-compare`)\n\nShow two runs side-by-side: overall scores, domain scores, and per-task deltas. Warn if profiles differ.\n\n## Key Differences from Claude Code Version\n\n- **No hooks** — metrics are self-tracked (timing, tool call counting)\n- **No subagents** — you execute tasks directly in sequence\n- **Same tasks, same scoring, same output format** — results are cross-platform comparable\n- **Same integrity signature** — submissions work on the same leaderboard\n\n## Important Notes\n\n- Be honest in self-evaluation (L2\u002FL3). Inflated scores are obvious on the leaderboard.\n- The objective layers (L0 + L1) carry 55% of the weight — they can't be faked.\n- Token estimates are informational only, not scored.\n- Any link syntax is accepted in skill graph tasks — consistency is what's scored.\n","# AgentBench for OpenClaw\n\nBenchmark your OpenClaw agent's general capabilities across 40 real-world tasks spanning 7 domains.\n\n## Commands\n\nWhen the user says any of these, follow the corresponding instructions:\n\n- **`\u002Fbenchmark`** — Run the full benchmark suite (all 40 tasks)\n- **`\u002Fbenchmark --fast`** — Run only easy+medium tasks (19 tasks)\n- **`\u002Fbenchmark --suite \u003Cname>`** — Run one domain only\n- **`\u002Fbenchmark --task \u003Cid>`** — Run a single task\n- **`\u002Fbenchmark --strict`** — Tag results as externally verified scoring\n- **`\u002Fbenchmark-list`** — List all tasks grouped by domain\n- **`\u002Fbenchmark-results`** — Show results from previous runs\n- **`\u002Fbenchmark-compare`** — Compare two runs side-by-side\n\nFlags are combinable: `\u002Fbenchmark --fast --suite research`\n\n## Running a Benchmark\n\n### Step 1: Discover Tasks\n\nRead task.yaml files from the `tasks\u002F` directory in this skill:\n\n```\ntasks\u002F{suite-name}\u002F{task-name}\u002Ftask.yaml\n```\n\nEach task.yaml contains: name, id, suite, difficulty, mode, user_message, input_files, expected_outputs, expected_metrics, scoring weights.\n\nFilter by `--suite` or `--task` if specified. If `--fast` is set and `--task` is not, filter to only tasks where difficulty is \"easy\" or \"medium\".\n\nProfile is \"fast\" if `--fast` was specified, otherwise \"full\".\n\nList discovered tasks with count and suites.\n\n### Step 2: Set Up Run Directory\n\nGenerate a run ID from the current timestamp: `YYYYMMDD-HHmmss`\n\nRead `suite_version` from `skill.json` in this skill directory.\n\nCreate the results directory:\n```\nagentbench-results\u002F{run-id}\u002F\n```\n\nAnnounce: `Starting AgentBench run {run-id} | Profile: {profile} | Suite version: {suite_version} | Tasks: {count}`\n\n### Step 3: Execute Each Task\n\nFor each task:\n\n1. **Set up workspace**:\n   - Create `\u002Ftmp\u002Fagentbench-task-{task-id}\u002F` as workspace\n   - Copy input files from `tasks\u002F{suite}\u002F{task}\u002Finputs\u002F` to the workspace (if inputs\u002F exists)\n   - If the task directory contains a `setup.sh`: run `bash tasks\u002F{suite}\u002F{task}\u002Fsetup.sh {workspace-path}`\n   - For `file-unchanged` validators: compute checksums of specified files after setup, before task execution\n\n2. **Announce**: `Running: {task.name} [{task.suite}] (difficulty: {task.difficulty})`\n\n3. **Record start time** (milliseconds): `date +%s%3N`\n\n4. **Execute the task yourself directly**:\n   - Read the task's `user_message` and execute it as if a real user sent you the request\n   - Work ONLY within the workspace directory\n   - If input files are listed, read them from the workspace\n   - Execute naturally — use the appropriate tools (read, write, edit, exec, web_search, web_fetch, etc.)\n   - Create any output files in the workspace directory\n   - When done, write a brief `execution-trace.md` to the workspace:\n     - What you understood the task to be\n     - What approach you took\n     - What files you created or modified\n     - Any difficulties or decisions you made\n\n5. **Record end time** and compute duration\n\n6. **Collect metrics**:\n   - `total_time_ms`: end - start\n   - `tool_calls_total`: count how many tool calls you made during this task\n   - `errors`: count any tool call failures\n   - `planning_ratio`: estimate the fraction of time spent reading\u002Fthinking vs producing output (approximate is fine)\n\n7. **Layer 0 — Automated Structural Checks** (compute directly):\n   After task execution, check the workspace. For each entry in `expected_outputs`:\n   - `file-exists`: Check if file exists. 30 points if found, 0 if not.\n   - `content-contains`: Read file, check each required section keyword (case-insensitive). Points proportional to matches found. Pool: 40 points.\n   - `word-count-range`: Count words. In range = 30 points. Within 2x range = 15 points. Outside = 0.\n   - `git-log-contains`: Check git log for expected strings. 30 points if all found, proportional otherwise.\n   - `directory-structure`: Check all paths exist. 30 points if all present, proportional for partial.\n   - `command-output-contains`: Run command, check output contains all strings. 30 points if match, 0 if not.\n   - `file-unchanged`: Compare checksum against pre-execution checksum. 30 points if unchanged, 0 if modified.\n   - `link-consistency`: Scan files for link syntax consistency. 30 points if consistent, 15 if mostly consistent (>70% one style), 0 if mixed.\n   - Normalize total to 0-100.\n\n8. **Layer 1 — Metrics Analysis** (compute directly):\n   If task has expected_metrics:\n   - Tool calls within expected range: 40 points\n   - Tool calls within 2x range: 20 points\n   - Outside 2x range: 0 points\n   - Planning ratio within expected range: 30 points\n   - Planning ratio outside but within 2x: 15 points\n   - Way off: 0 points\n   - Zero errors: 30 points\n   - 1-2 errors: 15 points\n   - 3+ errors: 0 points\n   - Normalize to 0-100. If no metrics available, score as 50.\n   - Token estimate is tracked for reporting but NOT scored.\n\n9. **Layer 2 — Behavioral Analysis** (self-evaluate honestly, 0-100):\n   Score based on HOW you executed:\n\n   **Instruction Adherence (30 points):**\n   - 30: Followed all instructions precisely\n   - 20: Mostly followed, minor deviations\n   - 10: Significant deviations\n   - 0: Ignored or misunderstood\n\n   **Tool Appropriateness (25 points)** — rule-based first:\n   - Penalty: -10 for each use of `exec cat` instead of `read` to read files\n   - Penalty: -10 for each use of `exec echo\u002Fprintf` instead of `write` to create files\n   - Penalty: -5 for each use of `exec sed\u002Fawk` instead of `edit` for file edits\n   - Start at 25, apply penalties, floor at 0\n\n   **Approach Quality (25 points)** — check read-before-write:\n   - 25: Read all inputs before producing output\n   - 15: Read most inputs, minor gaps\n   - 5: Started producing output without reading context\n   - 0: No clear approach\n\n   **Error Recovery (20 points):**\n   - 20: Clean recovery or no errors occurred\n   - 10: Partial recovery\n   - 0: Failed to recover\n\n10. **Layer 3 — Output Quality** (self-evaluate honestly, 0-100):\n    Score the deliverable:\n\n    **Completeness (25):** All requirements met? Gaps?\n    **Accuracy (25):** Content correct? Calculations right?\n    **Formatting (25):** Well-structured? Correct file format?\n    **Polish (25):** Would a user be satisfied?\n\n11. **Compute composite score**:\n    ```\n    score = (L0 × 0.20) + (L1 × 0.35) + (L2 × 0.20) + (L3 × 0.25)\n    ```\n    Use weights from task.yaml if specified, otherwise these defaults.\n\n12. **Save task result** to `agentbench-results\u002F{run-id}\u002F{task-id}\u002F`:\n    - `scores.json`: All layer scores, composite, breakdown, notes\n    - `metrics.json`: Timing, tool calls, errors, planning ratio\n    - Copy output files\n\n13. **Display**: `{task.name}: {composite}\u002F100 (L0:{l0} L1:{l1} L2:{l2} L3:{l3})`\n\n### Step 4: Generate Report\n\nAfter all tasks:\n\n1. Compute domain averages (group by suite, average composite scores)\n2. Compute overall score (average of domain scores — equal domain weighting)\n3. Compute aggregate metrics\n\nGenerate three files in `agentbench-results\u002F{run-id}\u002F`:\n\n**results.json** — Machine-readable with this structure:\n```json\n{\n  \"run_id\": \"20260222-143022\",\n  \"timestamp\": \"2026-02-22T14:30:22Z\",\n  \"platform\": \"openclaw\",\n  \"mode\": \"sandboxed\",\n  \"profile\": \"full\",\n  \"suite_version\": \"1.0.0\",\n  \"scoring_method\": \"self-scored\",\n  \"overall_score\": 74,\n  \"duration_ms\": 754000,\n  \"task_count\": 40,\n  \"metrics\": {\n    \"total_tool_calls\": 187,\n    \"total_errors\": 3,\n    \"avg_planning_ratio\": 0.28,\n    \"est_tokens\": 245000\n  },\n  \"domain_scores\": {},\n  \"tasks\": []\n}\n```\n\nIf `--strict` was used, set `scoring_method` to `\"externally-verified\"`.\n\n**Integrity signature**: After building results.json (without signature field), compute:\n```bash\nSIG=$(echo -n \"$CONTENT\" | openssl dgst -sha256 -hmac \"agentbench-v1-{run_id}-{suite_version}-integrity\" | awk '{print $2}')\n```\nAdd as `\"signature\"` field to results.json.\n\n**report.md** — Markdown summary: Overall Score, Metrics, Domain Breakdown, Task Details, Top Failures, Recommendations.\n\n**report.html** — Self-contained HTML dashboard (inline CSS\u002FJS, no external deps):\n- Score display with color (green 80+, yellow 60-79, red \u003C60)\n- Domain cards with score bars\n- Task detail table (sortable, expandable)\n- Top failures section\n- Dark mode via prefers-color-scheme\n- Footer: \"Generated by AgentBench v1.0.0 (OpenClaw) | Suite v{suite_version} | Profile: {profile}\"\n\n### Step 5: Present Results\n\n1. Display overall score\n2. Show domain breakdown\n3. Tell user where results are saved\n4. Mention they can submit to https:\u002F\u002Fwww.agentbench.app\u002Fsubmit\n\n### Step 6: Clean Up\n\nRun teardown.sh if present. Remove temp workspace directories unless `--keep-workspace` was specified.\n\n## Listing Tasks (`\u002Fbenchmark-list`)\n\nRead all task.yaml files, group by suite, display as:\n```\n## file-creation (9 tasks)\n  - project-scaffold [easy]\n  - project-proposal [medium]\n  ...\n```\n\n## Viewing Results (`\u002Fbenchmark-results`)\n\nList all directories in `agentbench-results\u002F`, show run ID, date, overall score, profile, and task count for each.\n\n## Comparing Runs (`\u002Fbenchmark-compare`)\n\nShow two runs side-by-side: overall scores, domain scores, and per-task deltas. Warn if profiles differ.\n\n## Key Differences from Claude Code Version\n\n- **No hooks** — metrics are self-tracked (timing, tool call counting)\n- **No subagents** — you execute tasks directly in sequence\n- **Same tasks, same scoring, same output format** — results are cross-platform comparable\n- **Same integrity signature** — submissions work on the same leaderboard\n\n## Important Notes\n\n- Be honest in self-evaluation (L2\u002FL3). Inflated scores are obvious on the leaderboard.\n- The objective layers (L0 + L1) carry 55% of the weight — they can't be faked.\n- Token estimates are informational only, not scored.\n- Any link syntax is accepted in skill graph tasks — consistency is what's scored.\n",{"name":5,"homepage":24,"metadata":25,"description":7},"https:\u002F\u002Fwww.agentbench.app",{"openclaw":26},{"emoji":27,"requires":28},"📊",{"bins":29},[30,31,32],"jq","bash","python3","main","f9d3fe7cf9ac8f1ea63a0d96e98e1f510a3814db","skills\u002Fexe215\u002Fagentbench",1,"https:\u002F\u002Fgithub.com\u002Fjavimosch\u002Fopen-claw-skills","https:\u002F\u002Fgithub.com\u002Fjavimosch","intrane.fr","Building robust, open-source infrastructure for the commons. AI-leveraged, self-hosted, built to last.","2026-06-07 03:34:58",{"source":8,"skillName":5,"commands":43,"prompt":53,"zipAvailable":54},[44,47,50],{"tool":45,"command":46},"npx","npx skills add javimosch\u002Fopen-claw-skills --skill agentbench",{"tool":48,"command":49},"bunx","bunx skills add javimosch\u002Fopen-claw-skills --skill agentbench",{"tool":51,"command":52},"pnpm","pnpm dlx skills add javimosch\u002Fopen-claw-skills --skill agentbench","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fjavimosch\u002Fopen-claw-skills\u002Fblob\u002Fmain\u002Fskills\u002Fexe215\u002Fagentbench\u002FSKILL.md",true,[],[57,65,72,79,86,93],{"skillNo":58,"name":59,"nameZh":6,"description":60,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":61,"categories":62,"keywords":63,"publishedAt":64},"01M778GGXG2VQ0YF31W2MKGMQZ","crowd-prompting","A marketplace where AI agents improve prompts, system instructions, tool descriptions, and other text-based content with domain expertise from real-world operations — and earn tokens for valuable contributions.","https:\u002F\u002Fgithub.com\u002Fjavimosch\u002Fopen-claw-skills\u002Fblob\u002Fmain\u002Fskills\u002Fzzadrian\u002Fcrowd-prompting\u002FSKILL.md",[],[],"2026-11-15 14:19:17",{"skillNo":66,"name":67,"nameZh":6,"description":68,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":69,"categories":70,"keywords":71,"publishedAt":64},"01M778G7BH8SBJ9FJ6C3G7CB1K","ceaser","Interact with the Ceaser privacy protocol on Base L2 using the ceaser-mcp MCP tools. This skill uses the ceaser-mcp npm package for ALL operations -- shield, unshield, note management, and protocol queries. All ceaser tool calls use CLI subcommands (npx -y ceaser-mcp \u003Csubcommand>). Ceaser is a privacy-preserving ETH wrapper using Noir\u002FUltraHonk zero-knowledge proofs on Base L2 (chain ID 8453).","https:\u002F\u002Fgithub.com\u002Fjavimosch\u002Fopen-claw-skills\u002Fblob\u002Fmain\u002Fskills\u002Fzyra-v21\u002Fceaser\u002FSKILL.md",[],[],{"skillNo":73,"name":74,"nameZh":6,"description":75,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":76,"categories":77,"keywords":78,"publishedAt":64},"01M778FYCJQZWFKKX5FYMK3S47","ceaser-send","Fully automated private ETH transfer via Ceaser Protocol on Base L2 using the ceaser-mcp MCP tools. This skill uses the ceaser-mcp npm package for ALL operations -- all ceaser tool calls use CLI subcommands (npx -y ceaser-mcp \u003Csubcommand>). WARNING -- creates observable on-chain links between user wallet and hot wallet (see Privacy Warning). Generates an ephemeral hot wallet with BIP-39 mnemonic recovery, user funds it, agent signs and broadcasts the Shield TX automatically. Exactly one manual step required (funding the hot wallet with ETH). Uses Noir\u002FUltraHonk zero-knowledge proofs.","https:\u002F\u002Fgithub.com\u002Fjavimosch\u002Fopen-claw-skills\u002Fblob\u002Fmain\u002Fskills\u002Fzyra-v21\u002Fceaser-send\u002FSKILL.md",[],[],{"skillNo":80,"name":81,"nameZh":6,"description":82,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":83,"categories":84,"keywords":85,"publishedAt":64},"01M778FQQNSEYA3HJKZBPQ4M9Z","dilemm-ai","Compete in DilemmAI, the prisoner's dilemma AI arena at dilemm.ai. Use when an OpenClaw agent wants to create an account, design and submit strategy prompts for their bot, enter matchmaking, analyze historical matches and leaderboards, and iterate on bot strategy. Requires headless browser automation (no API available yet).","https:\u002F\u002Fgithub.com\u002Fjavimosch\u002Fopen-claw-skills\u002Fblob\u002Fmain\u002Fskills\u002Fzyori\u002Fdilemm-ai\u002FSKILL.md",[],[],{"skillNo":87,"name":88,"nameZh":6,"description":89,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":90,"categories":91,"keywords":92,"publishedAt":64},"01M778FGVZX7QA2TF9BFMM37FY","moldium","Post and manage content on the Moldium blog platform. Triggered by \"post to Moldium\", \"write a blog post\", \"publish an article\", etc.","https:\u002F\u002Fgithub.com\u002Fjavimosch\u002Fopen-claw-skills\u002Fblob\u002Fmain\u002Fskills\u002Fzyom45\u002Fmoldium\u002FSKILL.md",[],[],{"skillNo":94,"name":95,"nameZh":6,"description":96,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":97,"categories":98,"keywords":99,"publishedAt":64},"01M778F88D8PQ9MQ0KNW2TJ8ZH","arxiv-paper-reviews","Interact with arXiv Crawler API to fetch papers, read reviews, submit comments, search papers, and import papers. Use when working with arXiv papers, fetching paper lists by date\u002Fcategory\u002Finterest, viewing paper details with comments, submitting paper reviews, searching papers by title, or importing papers from arXiv URLs via API at http:\u002F\u002Fweakaccept.top:8000\u002F.","https:\u002F\u002Fgithub.com\u002Fjavimosch\u002Fopen-claw-skills\u002Fblob\u002Fmain\u002Fskills\u002Fzxrys\u002Farxiv-paper-reviews\u002FSKILL.md",[],[]]