{ "schemaVersion": "1.0", "item": { "slug": "peer-review", "name": "Peer Review", "source": "tencent", "type": "skill", "category": "开发工具", "sourceUrl": "https://clawhub.ai/staybased/peer-review", "canonicalUrl": "https://clawhub.ai/staybased/peer-review", "targetPlatform": "OpenClaw" }, "install": { "downloadMode": "redirect", "downloadUrl": "/downloads/peer-review", "sourceDownloadUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=peer-review", "sourcePlatform": "tencent", "targetPlatform": "OpenClaw", "installMethod": "Manual import", "extraction": "Extract archive", "prerequisites": [ "OpenClaw" ], "packageFormat": "ZIP package", "includedAssets": [ "SKILL.md" ], "primaryDoc": "SKILL.md", "quickSetup": [ "Download the package from Yavira.", "Extract the archive and review SKILL.md first.", "Import or place the package into your OpenClaw setup." ], "agentAssist": { "summary": "Hand the extracted package to your coding agent with a concrete install brief instead of figuring it out manually.", "steps": [ "Download the package from Yavira.", "Extract it into a folder your agent can access.", "Paste one of the prompts below and point your agent at the extracted folder." ], "prompts": [ { "label": "New install", "body": "I downloaded a skill package from Yavira. Read SKILL.md from the extracted folder and install it by following the included instructions. Tell me what you changed and call out any manual steps you could not complete." }, { "label": "Upgrade existing", "body": "I downloaded an updated skill package from Yavira. Read SKILL.md from the extracted folder, compare it with my current installation, and upgrade it while preserving any custom configuration unless the package docs explicitly say otherwise. Summarize what changed and any follow-up checks I should run." } ] }, "sourceHealth": { "source": "tencent", "status": "healthy", "reason": "direct_download_ok", "recommendedAction": "download", "checkedAt": "2026-04-30T16:55:25.780Z", "expiresAt": "2026-05-07T16:55:25.780Z", "httpStatus": 200, "finalUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=network", "contentType": "application/zip", "probeMethod": "head", "details": { "probeUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=network", "contentDisposition": "attachment; filename=\"network-1.0.0.zip\"", "redirectLocation": null, "bodySnippet": null }, "scope": "source", "summary": "Source download looks usable.", "detail": "Yavira can redirect you to the upstream package for this source.", "primaryActionLabel": "Download for OpenClaw", "primaryActionHref": "/downloads/peer-review" }, "validation": { "installChecklist": [ "Use the Yavira download entry.", "Review SKILL.md after the package is downloaded.", "Confirm the extracted package contains the expected setup assets." ], "postInstallChecks": [ "Confirm the extracted package includes the expected docs or setup files.", "Validate the skill or prompts are available in your target agent workspace.", "Capture any manual follow-up steps the agent could not complete." ] }, "downloadPageUrl": "https://openagent3.xyz/downloads/peer-review", "agentPageUrl": "https://openagent3.xyz/skills/peer-review/agent", "manifestUrl": "https://openagent3.xyz/skills/peer-review/agent.json", "briefUrl": "https://openagent3.xyz/skills/peer-review/agent.md" }, "agentAssist": { "summary": "Hand the extracted package to your coding agent with a concrete install brief instead of figuring it out manually.", "steps": [ "Download the package from Yavira.", "Extract it into a folder your agent can access.", "Paste one of the prompts below and point your agent at the extracted folder." ], "prompts": [ { "label": "New install", "body": "I downloaded a skill package from Yavira. Read SKILL.md from the extracted folder and install it by following the included instructions. Tell me what you changed and call out any manual steps you could not complete." }, { "label": "Upgrade existing", "body": "I downloaded an updated skill package from Yavira. Read SKILL.md from the extracted folder, compare it with my current installation, and upgrade it while preserving any custom configuration unless the package docs explicitly say otherwise. Summarize what changed and any follow-up checks I should run." } ] }, "documentation": { "source": "clawhub", "primaryDoc": "SKILL.md", "sections": [ { "title": "Peer Review — Local LLM Critique Layer", "body": "Hypothesis: Local LLMs can catch ≥30% of real errors in cloud output with <50% false positive rate." }, { "title": "Architecture", "body": "Cloud Model (Claude) produces analysis\n │\n ▼\n┌────────────────────────┐\n│ Peer Review Fan-Out │\n├────────────────────────┤\n│ Drift (Mistral 7B) │──► Critique A\n│ Pip (TinyLlama 1.1B) │──► Critique B\n│ Lume (Llama 3.1 8B) │──► Critique C\n└────────────────────────┘\n │\n ▼\n Aggregator (consensus logic)\n │\n ▼\n Final: original + flagged issues" }, { "title": "Swarm Bot Roles", "body": "BotModelRoleStrengthsDrift 🌊Mistral 7BMethodical analystStructured reasoning, catches logical gapsPip 🐣TinyLlama 1.1BFast checkerQuick sanity checks, low latencyLume 💡Llama 3.1 8BDeep thinkerNuanced analysis, catches subtle issues" }, { "title": "Scripts", "body": "ScriptPurposescripts/peer-review.shSend single input to all models, collect critiquesscripts/peer-review-batch.shRun peer review across a corpus of samplesscripts/seed-test-corpus.shGenerate seeded error corpus for testing" }, { "title": "Usage", "body": "# Single file review\nbash scripts/peer-review.sh [output_dir]\n\n# Batch review\nbash scripts/peer-review-batch.sh [results_dir]\n\n# Generate test corpus\nbash scripts/seed-test-corpus.sh [count] [output_dir]\n\nScripts live at workspace/scripts/ — not bundled in skill to avoid duplication." }, { "title": "Critique Prompt Template", "body": "You are a skeptical reviewer. Analyze the following text for errors.\n\nFor each issue found, output JSON:\n{\"category\": \"factual|logical|missing|overconfidence|hallucinated_source\",\n \"quote\": \"...\", \"issue\": \"...\", \"confidence\": 0-100}\n\nIf no issues found, output: {\"issues\": []}\n\nTEXT:\n---\n{cloud_output}\n---" }, { "title": "Error Categories", "body": "CategoryDescriptionExamplefactualWrong numbers, dates, names\"Bitcoin launched in 2010\"logicalNon-sequiturs, unsupported conclusions\"X is rising, therefore Y will fall\"missingImportant context omittedIgnoring a major counterargumentoverconfidenceCertainty without justification\"This will definitely happen\" on 55% eventhallucinated_sourceCiting nonexistent sources\"According to a 2024 Reuters report...\"" }, { "title": "Discord Workflow", "body": "Post analysis to #the-deep (or #swarm-lab)\nDrift, Pip, and Lume respond with independent critiques\nCeleste synthesizes: deduplicates flags, weights by model confidence\nIf consensus (≥2 models agree) → flag is high-confidence\nFinal output posted with recommendation: publish | revise | flag_for_human" }, { "title": "Success Criteria", "body": "OutcomeTPRFPRDecisionStrong pass≥50%<30%Ship as default layerPass≥30%<50%Ship as opt-in layerMarginal20–30%50–70%Iterate on prompts, retestFail<20%>70%Abandon approach" }, { "title": "Scoring Rules", "body": "Flag = true positive if it identifies a real error (even if explanation is imperfect)\nFlag = false positive if flagged content is actually correct\nDuplicate flags across models count once for TPR but inform consensus metrics" }, { "title": "Dependencies", "body": "Ollama running locally with models pulled: mistral:7b, tinyllama:1.1b, llama3.1:8b\njq and curl installed\nResults stored in experiments/peer-review-results/" }, { "title": "Integration", "body": "When peer review passes validation:\n\nPackage as Reef API endpoint: POST /review\nAgents call before publishing any analysis\nConfigurable: model selection, consensus threshold, categories\nLog all reviews to #reef-logs with TPR tracking" } ], "body": "Peer Review — Local LLM Critique Layer\n\nHypothesis: Local LLMs can catch ≥30% of real errors in cloud output with <50% false positive rate.\n\nArchitecture\nCloud Model (Claude) produces analysis\n │\n ▼\n┌────────────────────────┐\n│ Peer Review Fan-Out │\n├────────────────────────┤\n│ Drift (Mistral 7B) │──► Critique A\n│ Pip (TinyLlama 1.1B) │──► Critique B\n│ Lume (Llama 3.1 8B) │──► Critique C\n└────────────────────────┘\n │\n ▼\n Aggregator (consensus logic)\n │\n ▼\n Final: original + flagged issues\n\nSwarm Bot Roles\nBot\tModel\tRole\tStrengths\nDrift 🌊\tMistral 7B\tMethodical analyst\tStructured reasoning, catches logical gaps\nPip 🐣\tTinyLlama 1.1B\tFast checker\tQuick sanity checks, low latency\nLume 💡\tLlama 3.1 8B\tDeep thinker\tNuanced analysis, catches subtle issues\nScripts\nScript\tPurpose\nscripts/peer-review.sh\tSend single input to all models, collect critiques\nscripts/peer-review-batch.sh\tRun peer review across a corpus of samples\nscripts/seed-test-corpus.sh\tGenerate seeded error corpus for testing\nUsage\n# Single file review\nbash scripts/peer-review.sh [output_dir]\n\n# Batch review\nbash scripts/peer-review-batch.sh [results_dir]\n\n# Generate test corpus\nbash scripts/seed-test-corpus.sh [count] [output_dir]\n\n\nScripts live at workspace/scripts/ — not bundled in skill to avoid duplication.\n\nCritique Prompt Template\nYou are a skeptical reviewer. Analyze the following text for errors.\n\nFor each issue found, output JSON:\n{\"category\": \"factual|logical|missing|overconfidence|hallucinated_source\",\n \"quote\": \"...\", \"issue\": \"...\", \"confidence\": 0-100}\n\nIf no issues found, output: {\"issues\": []}\n\nTEXT:\n---\n{cloud_output}\n---\n\nError Categories\nCategory\tDescription\tExample\nfactual\tWrong numbers, dates, names\t\"Bitcoin launched in 2010\"\nlogical\tNon-sequiturs, unsupported conclusions\t\"X is rising, therefore Y will fall\"\nmissing\tImportant context omitted\tIgnoring a major counterargument\noverconfidence\tCertainty without justification\t\"This will definitely happen\" on 55% event\nhallucinated_source\tCiting nonexistent sources\t\"According to a 2024 Reuters report...\"\nDiscord Workflow\nPost analysis to #the-deep (or #swarm-lab)\nDrift, Pip, and Lume respond with independent critiques\nCeleste synthesizes: deduplicates flags, weights by model confidence\nIf consensus (≥2 models agree) → flag is high-confidence\nFinal output posted with recommendation: publish | revise | flag_for_human\nSuccess Criteria\nOutcome\tTPR\tFPR\tDecision\nStrong pass\t≥50%\t<30%\tShip as default layer\nPass\t≥30%\t<50%\tShip as opt-in layer\nMarginal\t20–30%\t50–70%\tIterate on prompts, retest\nFail\t<20%\t>70%\tAbandon approach\nScoring Rules\nFlag = true positive if it identifies a real error (even if explanation is imperfect)\nFlag = false positive if flagged content is actually correct\nDuplicate flags across models count once for TPR but inform consensus metrics\nDependencies\nOllama running locally with models pulled: mistral:7b, tinyllama:1.1b, llama3.1:8b\njq and curl installed\nResults stored in experiments/peer-review-results/\nIntegration\n\nWhen peer review passes validation:\n\nPackage as Reef API endpoint: POST /review\nAgents call before publishing any analysis\nConfigurable: model selection, consensus threshold, categories\nLog all reviews to #reef-logs with TPR tracking" }, "trust": { "sourceLabel": "tencent", "provenanceUrl": "https://clawhub.ai/staybased/peer-review", "publisherUrl": "https://clawhub.ai/staybased/peer-review", "owner": "staybased", "version": "1.0.0", "license": null, "verificationStatus": "Indexed source record" }, "links": { "detailUrl": "https://openagent3.xyz/skills/peer-review", "downloadUrl": "https://openagent3.xyz/downloads/peer-review", "agentUrl": "https://openagent3.xyz/skills/peer-review/agent", "manifestUrl": "https://openagent3.xyz/skills/peer-review/agent.json", "briefUrl": "https://openagent3.xyz/skills/peer-review/agent.md" } }