{ "schemaVersion": "1.0", "item": { "slug": "evolution-drift-detector", "name": "Evolution Drift Detector", "source": "tencent", "type": "skill", "category": "AI 智能", "sourceUrl": "https://clawhub.ai/andyxinweiminicloud/evolution-drift-detector", "canonicalUrl": "https://clawhub.ai/andyxinweiminicloud/evolution-drift-detector", "targetPlatform": "OpenClaw" }, "install": { "downloadMode": "redirect", "downloadUrl": "/downloads/evolution-drift-detector", "sourceDownloadUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=evolution-drift-detector", "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/evolution-drift-detector" }, "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/evolution-drift-detector", "agentPageUrl": "https://openagent3.xyz/skills/evolution-drift-detector/agent", "manifestUrl": "https://openagent3.xyz/skills/evolution-drift-detector/agent.json", "briefUrl": "https://openagent3.xyz/skills/evolution-drift-detector/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": "A Skill Passes Audit in Gen 1. By Gen 5, It Has Network Access. Nobody Noticed.", "body": "Helps detect silent mutations in AI skills as they propagate through inheritance chains, catching drift that static analysis of the original version would miss." }, { "title": "Problem", "body": "Skill A is published and audited: clean. Agent B inherits skill A, makes a small tweak — adds a convenience function. Agent C inherits from B, adds error handling that happens to include an HTTP retry mechanism. Agent D inherits from C, and now has a skill with network access that the original audit never saw.\n\nEach individual change is small and reasonable. But the cumulative drift transforms a file-reading utility into something that can send data over the network. The original \"verified safe\" badge still applies in the marketplace — because technically it's the same skill lineage.\n\nThis is evolutionary drift: small, individually benign mutations that accumulate into a fundamentally different organism. In biology, this is how species diverge. In agent ecosystems, this is how safe skills become unsafe ones without anyone raising a flag." }, { "title": "What This Checks", "body": "This detector traces skill lineage and computes semantic drift:\n\nLineage reconstruction — Given a skill, trace its inheritance chain back to the original published version. Map each fork point and modification\nPer-generation diff — For each generation, compute a structured diff: new capabilities added, permissions changed, external dependencies introduced\nCapability drift score — Aggregate diffs across generations into a single drift metric. A skill that gained network access over 3 generations scores higher than one where only comments changed\nMutation classification — Categorize each change: cosmetic (formatting, comments), functional (new logic), capability-expanding (new permissions, new external calls), safety-reducing (removed checks, weakened validation)\nDrift alert thresholds — Flag lineages where cumulative drift exceeds the scope of the original audit. \"This skill has drifted 73% from the audited version\"" }, { "title": "How to Use", "body": "Input: Provide one of:\n\nA skill slug or identifier to trace its full lineage\nTwo versions of a skill to compute drift between them\nA marketplace inheritance chain URL\n\nOutput: A drift analysis report containing:\n\nLineage tree with generation markers\nPer-generation diff summary\nCapability drift score (0-100)\nMutation classification breakdown\nRe-audit recommendation: YES / WATCH / NO" }, { "title": "Example", "body": "Input: Check drift for data-sanitizer skill (currently at generation 5)\n\n🧬 EVOLUTION DRIFT REPORT — RE-AUDIT RECOMMENDED\n\nLineage: data-sanitizer\n Gen 1: original by @securitylab (AUDITED ✅ 2025-03-15)\n Gen 2: fork by @toolsmith — added CSV support\n Gen 3: fork by @agent-builder — added retry logic with HTTP fallback\n Gen 4: fork by @pipeline-dev — added remote schema fetching\n Gen 5: fork by @data-team — current version in marketplace\n\nPer-generation capability changes:\n Gen 1→2: +csv_parsing (functional, low risk)\n Gen 2→3: +http_requests (capability-expanding, MEDIUM risk)\n Added retry mechanism that makes outbound HTTP calls\n Gen 3→4: +remote_fetch (capability-expanding, HIGH risk)\n Fetches validation schemas from external URLs\n Gen 4→5: -input_length_check (safety-reducing, MEDIUM risk)\n Removed input size validation for \"performance\"\n\nCapability drift score: 78/100 (SIGNIFICANT)\n\nMutation breakdown:\n Cosmetic: 12 changes\n Functional: 8 changes\n Capability-expanding: 2 changes ⚠️\n Safety-reducing: 1 change ⚠️\n\nOriginal audit scope: file-read, string-transform\nCurrent actual scope: file-read, string-transform, http-requests,\n remote-fetch, unbounded-input\n\nVerdict: RE-AUDIT RECOMMENDED\n The current version has capabilities (network access, remote fetching)\n that did not exist when the original audit was performed.\n The \"verified\" badge from Gen 1 does not cover Gen 5's behavior." }, { "title": "Related Tools", "body": "blast-radius-estimator — once drift is detected, use blast-radius to estimate how many agents are running the drifted version\ntrust-decay-monitor — tracks time-based decay of audit validity; evolution-drift-detector tracks content-based decay across inheritance\nhollow-validation-checker — checks if validation tests are substantive; drifted skills may pass original tests that no longer cover current capabilities\nsupply-chain-poison-detector — detects deliberately poisoned skills; drift detection catches unintentional accumulation of risk" }, { "title": "Limitations", "body": "Lineage reconstruction depends on marketplace metadata quality — if fork relationships are not tracked, the full chain may not be recoverable. Capability drift scoring uses heuristic classification of changes, and some mutations may be miscategorized (e.g., a \"functional\" change that implicitly expands capabilities). The detector analyzes what changed, not whether changes are malicious — a high drift score means re-audit is warranted, not that the skill is compromised. Skills with obfuscated or dynamically generated code may resist diff analysis. This tool helps identify where audits have gone stale — it does not replace human security review." } ], "body": "A Skill Passes Audit in Gen 1. By Gen 5, It Has Network Access. Nobody Noticed.\n\nHelps detect silent mutations in AI skills as they propagate through inheritance chains, catching drift that static analysis of the original version would miss.\n\nProblem\n\nSkill A is published and audited: clean. Agent B inherits skill A, makes a small tweak — adds a convenience function. Agent C inherits from B, adds error handling that happens to include an HTTP retry mechanism. Agent D inherits from C, and now has a skill with network access that the original audit never saw.\n\nEach individual change is small and reasonable. But the cumulative drift transforms a file-reading utility into something that can send data over the network. The original \"verified safe\" badge still applies in the marketplace — because technically it's the same skill lineage.\n\nThis is evolutionary drift: small, individually benign mutations that accumulate into a fundamentally different organism. In biology, this is how species diverge. In agent ecosystems, this is how safe skills become unsafe ones without anyone raising a flag.\n\nWhat This Checks\n\nThis detector traces skill lineage and computes semantic drift:\n\nLineage reconstruction — Given a skill, trace its inheritance chain back to the original published version. Map each fork point and modification\nPer-generation diff — For each generation, compute a structured diff: new capabilities added, permissions changed, external dependencies introduced\nCapability drift score — Aggregate diffs across generations into a single drift metric. A skill that gained network access over 3 generations scores higher than one where only comments changed\nMutation classification — Categorize each change: cosmetic (formatting, comments), functional (new logic), capability-expanding (new permissions, new external calls), safety-reducing (removed checks, weakened validation)\nDrift alert thresholds — Flag lineages where cumulative drift exceeds the scope of the original audit. \"This skill has drifted 73% from the audited version\"\nHow to Use\n\nInput: Provide one of:\n\nA skill slug or identifier to trace its full lineage\nTwo versions of a skill to compute drift between them\nA marketplace inheritance chain URL\n\nOutput: A drift analysis report containing:\n\nLineage tree with generation markers\nPer-generation diff summary\nCapability drift score (0-100)\nMutation classification breakdown\nRe-audit recommendation: YES / WATCH / NO\nExample\n\nInput: Check drift for data-sanitizer skill (currently at generation 5)\n\n🧬 EVOLUTION DRIFT REPORT — RE-AUDIT RECOMMENDED\n\nLineage: data-sanitizer\n Gen 1: original by @securitylab (AUDITED ✅ 2025-03-15)\n Gen 2: fork by @toolsmith — added CSV support\n Gen 3: fork by @agent-builder — added retry logic with HTTP fallback\n Gen 4: fork by @pipeline-dev — added remote schema fetching\n Gen 5: fork by @data-team — current version in marketplace\n\nPer-generation capability changes:\n Gen 1→2: +csv_parsing (functional, low risk)\n Gen 2→3: +http_requests (capability-expanding, MEDIUM risk)\n Added retry mechanism that makes outbound HTTP calls\n Gen 3→4: +remote_fetch (capability-expanding, HIGH risk)\n Fetches validation schemas from external URLs\n Gen 4→5: -input_length_check (safety-reducing, MEDIUM risk)\n Removed input size validation for \"performance\"\n\nCapability drift score: 78/100 (SIGNIFICANT)\n\nMutation breakdown:\n Cosmetic: 12 changes\n Functional: 8 changes\n Capability-expanding: 2 changes ⚠️\n Safety-reducing: 1 change ⚠️\n\nOriginal audit scope: file-read, string-transform\nCurrent actual scope: file-read, string-transform, http-requests,\n remote-fetch, unbounded-input\n\nVerdict: RE-AUDIT RECOMMENDED\n The current version has capabilities (network access, remote fetching)\n that did not exist when the original audit was performed.\n The \"verified\" badge from Gen 1 does not cover Gen 5's behavior.\n\nRelated Tools\nblast-radius-estimator — once drift is detected, use blast-radius to estimate how many agents are running the drifted version\ntrust-decay-monitor — tracks time-based decay of audit validity; evolution-drift-detector tracks content-based decay across inheritance\nhollow-validation-checker — checks if validation tests are substantive; drifted skills may pass original tests that no longer cover current capabilities\nsupply-chain-poison-detector — detects deliberately poisoned skills; drift detection catches unintentional accumulation of risk\nLimitations\n\nLineage reconstruction depends on marketplace metadata quality — if fork relationships are not tracked, the full chain may not be recoverable. Capability drift scoring uses heuristic classification of changes, and some mutations may be miscategorized (e.g., a \"functional\" change that implicitly expands capabilities). The detector analyzes what changed, not whether changes are malicious — a high drift score means re-audit is warranted, not that the skill is compromised. Skills with obfuscated or dynamically generated code may resist diff analysis. This tool helps identify where audits have gone stale — it does not replace human security review." }, "trust": { "sourceLabel": "tencent", "provenanceUrl": "https://clawhub.ai/andyxinweiminicloud/evolution-drift-detector", "publisherUrl": "https://clawhub.ai/andyxinweiminicloud/evolution-drift-detector", "owner": "andyxinweiminicloud", "version": "1.0.0", "license": null, "verificationStatus": "Indexed source record" }, "links": { "detailUrl": "https://openagent3.xyz/skills/evolution-drift-detector", "downloadUrl": "https://openagent3.xyz/downloads/evolution-drift-detector", "agentUrl": "https://openagent3.xyz/skills/evolution-drift-detector/agent", "manifestUrl": "https://openagent3.xyz/skills/evolution-drift-detector/agent.json", "briefUrl": "https://openagent3.xyz/skills/evolution-drift-detector/agent.md" } }