[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7AYT5TA1CCRPCDM758P98C9":3,"skill-same:01M7AYT5TA1CCRPCDM758P98C9":75},{"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":42,"commitSha":43,"skillDirectory":44,"forksCount":45,"repositoryHtmlUrl":46,"repositoryHomepageUrl":6,"ownerHtmlUrl":47,"ownerBlogUrl":48,"ownerDescription":49,"updatedAt":50,"install":51,"resources":64},"01M7AYT5TA1CCRPCDM758P98C9","evidence-based-document-audit",null,"Use when auditing document sets against standards.","oat431\u002Fhermes_config_backup","oat431","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F54722120?v=4","01M7AYD07EH8SYPMYY8CGZKS22","Oralita","01M7AYD0816QDF77E3TWCRSQC9",1,"MIT","https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fproduct-owner\u002Fevidence-based-document-audit\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: evidence-based-document-audit\ndescription: Use when auditing document sets against standards.\nversion: 1.0.0\nauthor: Hermes Agent\nlicense: MIT\nplatforms: [linux, macos, windows]\nmetadata:\n  hermes:\n    tags: [research, document-audit, evidence, standards, swebok, traceability]\n    related_skills: [privacy-regulatory-research, hermes-agent-skill-authoring]\n---\n\n# Evidence-Based Document-Set Audits\n\n## Overview\n\nUse this skill to audit a proposed “essential,” “required,” or “standard” document\u002Fartifact catalog against an authoritative body of knowledge, standards, and a local knowledge directory. The deliverable is an evidence-backed report, not a rewritten catalog: identify omissions, over\u002Funder-prioritized items, duplicates, and weak or incorrect attributions with file paths and line references.\n\nThe central distinction is **knowledge coverage versus mandated work products**. A body of knowledge normally describes concepts, processes, inputs, outputs, practices, and tools; it rarely mandates one universal set of standalone documents. Treat any “essential” label as a claim that must be tested against lifecycle context, system criticality, methodology, team scale, regulatory obligations, and whether the artifact can be represented in a tool rather than a separate file.\n\n## When to Use\n\nUse when the user asks to:\n\n- audit an artifact\u002Fdocument inventory against SWEBOK, a standard, a framework, a policy, or a local knowledge base;\n- find missing categories, incorrect priorities, duplicates, or unsupported citations;\n- review a “required documents” checklist without editing it;\n- produce a concise report with evidence and line-level references.\n\nDo not use this for ordinary proofreading, a source summary, or requirements elicitation unless the user is explicitly auditing a catalog.\n\n## Workflow\n\n### 1. Establish the audit boundary\n\nRecord the exact target file, authoritative source files, local-note directory, and whether edits are prohibited. Treat the target as read-only. Identify whether “document” includes code, tests, configuration, records, dashboards, and tool-managed data; if the catalog mixes them, call that out rather than silently normalizing it.\n\nCompletion criterion: the report can name every audited root and the target’s line range or table-entry count.\n\n### 2. Inventory the target structurally\n\nRead the target with line numbers. Extract section headings, table rows, labels, priorities, descriptions, and source\u002Fstandard columns. Count entries by priority. Normalize labels only for analysis; preserve the original wording and line references in the report.\n\nLook for:\n\n- rows that are activities, methods, tools, code, configurations, records, or reports rather than documents;\n- conditional labels such as Agile, OOP, safety-critical, regulated, or production;\n- repeated labels, parent\u002Fsubset relationships, and cross-cutting rows that subsume phase-specific rows;\n- claims that the catalog is “SWEBOK-derived” or “essential for virtually all projects.”\n\nCompletion criterion: every finding points to an exact target row or an explicitly documented absence in the relevant section.\n\n### 3. Build a source-to-artifact map\n\nRead the source overview first, then the relevant knowledge-area chapters and local detailed notes. Extract source concepts in four buckets:\n\n1. **Processes and activities** — e.g., requirements validation, change control, incident management.\n2. **Inputs and outputs** — e.g., baselines, test results, release packages, audit evidence.\n3. **Techniques and models** — e.g., ATAM, UML, FMEA, QAW, TDD.\n4. **Context qualifiers** — e.g., safety-critical, service operation, Agile, large system, formal CM.\n\nDo not infer a required standalone document merely because a source mentions an artifact. Verify whether the source says it is an output, a recommended record, a tool-managed item, or only an example.\n\nCompletion criterion: each major lifecycle\u002Fknowledge area has at least one source citation and a comparison against the target.\n\n### 4. Identify missing categories\n\nSearch for omissions in these common classes:\n\n- concept and feasibility: vision\u002Fscope, business case, feasibility, proposal\u002Fcontract;\n- planning\u002Fcontrol: scope baseline, estimate basis, cost\u002Fschedule baselines, status reports, decisions, measurement plan;\n- requirements governance: requirements baseline, backlog\u002Frelease allocation, issue\u002FTBD register, review record, CCB charter;\n- product delivery: user\u002Finstall\u002Ftraining\u002Fsupport documentation, deployment package, build manifest, release-readiness evidence;\n- testing and quality: test policy\u002Fstrategy, environment record, execution\u002Fincident log, summary report, V&V report, quality gates, CAPA, assurance case;\n- operations\u002Fsupport: operations plan, service reports, incident\u002Fpostmortem, problem\u002Fknown-error records, service desk, recovery rehearsal;\n- maintenance\u002Fretirement: transition plan, data migration, archive, decommissioning;\n- security: security plan, risk assessment\u002Facceptance, incident response, posture and dependency-compliance reports;\n- process and organizational improvement: process tailoring, measurement repository, retrospective\u002Flessons-learned, improvement plan.\n\nUse “missing as a named category” when an existing broader row may contain the content. Do not claim a concept is wholly absent if the source material or target contains it under another name.\n\n### 5. Audit priorities using context, not intuition\n\nStart from the target’s own priority definition. For each Must Have, ask:\n\n- Is the underlying activity essential, or only the separate document name?\n- Is it applicable to virtually all projects, or only to a system type\u002Flifecycle\u002Fdomain?\n- Can it be represented by an SRS, backlog, repository, issue tracker, CI system, or dashboard?\n- Does the source explicitly require it, or merely mention it as an example?\n\nFlag **over-classification** when a conditional artifact is marked universal: formal architecture packages, ADRs\u002F4+1 views, SBOMs, UAT sign-offs, SLAs, DR plans, FCA\u002FPCA, QMS, formal V&V, and security reports are common examples. Flag **under-classification** when a source calls something a core activity\u002Foutput but the catalog marks it Optional\u002FNice, such as requirements baselining, status accounting under formal CM, or recovery evidence for an in-scope production service.\n\nPrefer a three-axis recommendation when possible:\n\n- **Core practice:** generally expected activity;\n- **Artifact form:** separate document, record, tool-managed item, or generated report;\n- **Applicability:** universal, common, conditional, or domain-specific.\n\n### 6. Detect duplicates by relationship, not just exact text\n\nClassify overlaps as:\n\n- **Exact duplicate:** same concept and scope under two headings;\n- **Parent\u002Fsubset:** e.g., test suite contains test cases; SAST is a security subset of static analysis;\n- **Lifecycle variant:** same artifact at requirements, design, implementation, or maintenance scope;\n- **Format\u002Fcontent duplicate:** ADR and design rationale, release notes and VDD, API contract and API reference docs;\n- **Process versus record:** incident-management process versus incident record\u002Fpostmortem;\n- **Cross-cutting view:** one end-to-end RTM versus phase-specific traceability views.\n\nRecommend a single canonical category with scope\u002Ftype fields instead of deleting useful distinctions.\n\n### 7. Verify source and standard attributions\n\nFor each citation, check four things:\n\n1. **Scope:** does the standard govern this artifact or only a neighboring process?\n2. **Edition\u002Fstatus:** is the cited edition current, withdrawn, superseded, or unspecified?\n3. **Type:** is it an international standard, IEEE standard, framework, specification, taxonomy, regulation, tool, or book?\n4. **Local support:** does the cited source file actually mention or support the attribution?\n\nCommon errors include using a quality model as a report template, an ISMS standard as a SAST\u002Fsecurity-architecture standard, an architecture-description standard as an ADR-format standard, an open-source compliance standard as an SBOM standard, or a generic UML citation without a version. Separate columns for `Source\u002FKA`, `Artifact standard`, `Framework\u002Ftaxonomy`, and `External practice` are preferable.\n\nCompletion criterion: every “incorrect” attribution is supported either by the authoritative standard’s stated scope or by a clear mismatch between the target citation and the local source text.\n\n### 8. Write the report, not a replacement catalog\n\nUse this compact structure:\n\n1. **Scope and headline finding** — roots audited, entry count, major pattern.\n2. **Missing categories** — grouped by lifecycle\u002Fknowledge area, with source and target references.\n3. **Priority errors** — over-classified, under-classified, and context-dependent items.\n4. **Duplicates\u002Foverlaps** — original rows plus recommended relationship.\n5. **Source\u002Fstandard corrections** — precise citation issues and evidence.\n6. **Files changed** — explicitly say none when the audit was read-only.\n7. **Tooling issue** — mention only if it materially affected the method; describe the workaround, not a permanent refusal.\n\nKeep conclusions proportional to evidence. Use “not named,” “overlaps,” “weakly supported,” or “conditional” where absolute claims would overstate the comparison.\n\n## Common Pitfalls\n\n1. **Treating SWEBOK as a mandatory document template.** SWEBOK is a body-of-knowledge map; distinguish generally accepted practice from project-specific work-product choices.\n2. **Counting a separate file as the essential thing.** The essential thing may be traceability, change control, testing evidence, or operational knowledge, represented in a tool or combined document.\n3. **Using absence as proof without checking aliases.** Search for synonyms such as vision\u002Fscope versus BRD, status report versus dashboard, postmortem versus RCA, and deployment package versus release artifact.\n4. **Calling every overlap a duplicate.** Preserve useful lifecycle or scope distinctions; report parent\u002Fsubset and process\u002Frecord relationships.\n5. **Applying a standard outside its scope.** Verify whether it defines a process, a data format, a quality model, an ISMS, a review framework, or an artifact template.\n6. **Ignoring conditional rows.** Reclassify by applicability instead of forcing one universal priority.\n7. **Citing a local summary as if it were the primary standard.** Identify whether the note is a paraphrase, a book-derived practice, or the actual authoritative publication.\n8. **Replacing the requested audit with edits.** The user asked for evidence; do not modify the catalog unless explicitly requested.\n9. **Overstating current-edition facts.** Include the edition\u002Fyear and status when the claim depends on currency.\n10. **Trusting prior remediation claims.** When the target set contains an earlier audit or review doc marked \"✅ FIXED\", re-verify every claimed fix with a machine check (grep\u002Fscript) before accepting it — quick passes routinely claim blanket fixes that only partially landed, and repeating the claim launders a false fix into the next baseline.\n11. **Patching a self-contradictory index.** When a master checklist's totals disagree with each other AND with the files on disk, do not renumber it in place; flag it for regeneration-from-disk after the categorization decision, because hand-maintained counts drift again immediately.\n\n## Obsidian\u002FMarkdown Template-Set Audits\n\nWhen the target is an Obsidian vault or Markdown repository, apply the additional probes below:\n\n- Treat the folder name `document_template` as a claim to test: distinguish a document catalog\u002Fchecklist from actual reusable templates with metadata and lifecycle fields.\n- Resolve Obsidian wikilinks and literal backtick paths separately; a healthy wikilink graph does not prove that embedded filesystem paths are current. Strip fenced code blocks before wikilink resolution — shell `[[ ... ]]` test operators otherwise register as broken links — and remember placeholder text with nested brackets (`[[X]%]`) also parses as a wikilink, a recurring false-broken source in placeholder-ized templates.\n- Compare profile counts and Quick-Start checklists against the priority-marked rows. Report both true omissions and abbreviation\u002Fnaming mismatches.\n- Build an applicability matrix that separates core practice, artifact form, and context (universal, common, conditional, domain-specific, or not applicable).\n- Reconcile overview\u002Findex counts and scope claims across the source corpus before treating extracted labels as authoritative.\n- Check repository status before\u002Fafter a read-only audit so pre-existing working-tree changes are not attributed to the audit.\n- When the templates feed AI-persona-driven spec generation, scan every template for example-content contamination (concrete IDs, currency, dates, named example entities that a consumer could inherit as real data) and report per-file scores, not just spot-checks.\n- When the user asks for a large restructure, the deliverable is a dated plan doc + companion data doc written into the vault — never in-place edits to the audited templates during the audit session.\n\nSee `references\u002Fobsidian-template-audit.md` for a reusable probe sequence, evidence patterns, standards-verification examples, and the contamination-scan scoring recipe.\n\n## Verification Checklist\n\n- [ ] Target file, source roots, and read-only scope are stated.\n- [ ] Target entries are counted and priority totals are grounded in inspection.\n- [ ] Every major finding has a target path\u002Fline and a source path\u002Fline where possible.\n- [ ] Missing categories distinguish absent names from absent concepts.\n- [ ] Priority findings distinguish core practice from standalone artifact and applicability.\n- [ ] Duplicate findings identify exact, parent\u002Fsubset, lifecycle, or process\u002Frecord relationships.\n- [ ] Standard attributions are checked for scope, edition\u002Fstatus, and type.\n- [ ] External authoritative links are included only for disputed\u002Fcurrent standard facts.\n- [ ] No files were edited or created unless the user explicitly authorized it.\n- [ ] The final report is concise enough to scan but specific enough to reproduce.\n\n## Supporting References\n\n- See `references\u002Fswebok-artifact-audit.md` for a condensed source map, artifact-category checklist, and standard-attribution traps from a completed SWEBOK audit.\n","# Evidence-Based Document-Set Audits\n\n## Overview\n\nUse this skill to audit a proposed “essential,” “required,” or “standard” document\u002Fartifact catalog against an authoritative body of knowledge, standards, and a local knowledge directory. The deliverable is an evidence-backed report, not a rewritten catalog: identify omissions, over\u002Funder-prioritized items, duplicates, and weak or incorrect attributions with file paths and line references.\n\nThe central distinction is **knowledge coverage versus mandated work products**. A body of knowledge normally describes concepts, processes, inputs, outputs, practices, and tools; it rarely mandates one universal set of standalone documents. Treat any “essential” label as a claim that must be tested against lifecycle context, system criticality, methodology, team scale, regulatory obligations, and whether the artifact can be represented in a tool rather than a separate file.\n\n## When to Use\n\nUse when the user asks to:\n\n- audit an artifact\u002Fdocument inventory against SWEBOK, a standard, a framework, a policy, or a local knowledge base;\n- find missing categories, incorrect priorities, duplicates, or unsupported citations;\n- review a “required documents” checklist without editing it;\n- produce a concise report with evidence and line-level references.\n\nDo not use this for ordinary proofreading, a source summary, or requirements elicitation unless the user is explicitly auditing a catalog.\n\n## Workflow\n\n### 1. Establish the audit boundary\n\nRecord the exact target file, authoritative source files, local-note directory, and whether edits are prohibited. Treat the target as read-only. Identify whether “document” includes code, tests, configuration, records, dashboards, and tool-managed data; if the catalog mixes them, call that out rather than silently normalizing it.\n\nCompletion criterion: the report can name every audited root and the target’s line range or table-entry count.\n\n### 2. Inventory the target structurally\n\nRead the target with line numbers. Extract section headings, table rows, labels, priorities, descriptions, and source\u002Fstandard columns. Count entries by priority. Normalize labels only for analysis; preserve the original wording and line references in the report.\n\nLook for:\n\n- rows that are activities, methods, tools, code, configurations, records, or reports rather than documents;\n- conditional labels such as Agile, OOP, safety-critical, regulated, or production;\n- repeated labels, parent\u002Fsubset relationships, and cross-cutting rows that subsume phase-specific rows;\n- claims that the catalog is “SWEBOK-derived” or “essential for virtually all projects.”\n\nCompletion criterion: every finding points to an exact target row or an explicitly documented absence in the relevant section.\n\n### 3. Build a source-to-artifact map\n\nRead the source overview first, then the relevant knowledge-area chapters and local detailed notes. Extract source concepts in four buckets:\n\n1. **Processes and activities** — e.g., requirements validation, change control, incident management.\n2. **Inputs and outputs** — e.g., baselines, test results, release packages, audit evidence.\n3. **Techniques and models** — e.g., ATAM, UML, FMEA, QAW, TDD.\n4. **Context qualifiers** — e.g., safety-critical, service operation, Agile, large system, formal CM.\n\nDo not infer a required standalone document merely because a source mentions an artifact. Verify whether the source says it is an output, a recommended record, a tool-managed item, or only an example.\n\nCompletion criterion: each major lifecycle\u002Fknowledge area has at least one source citation and a comparison against the target.\n\n### 4. Identify missing categories\n\nSearch for omissions in these common classes:\n\n- concept and feasibility: vision\u002Fscope, business case, feasibility, proposal\u002Fcontract;\n- planning\u002Fcontrol: scope baseline, estimate basis, cost\u002Fschedule baselines, status reports, decisions, measurement plan;\n- requirements governance: requirements baseline, backlog\u002Frelease allocation, issue\u002FTBD register, review record, CCB charter;\n- product delivery: user\u002Finstall\u002Ftraining\u002Fsupport documentation, deployment package, build manifest, release-readiness evidence;\n- testing and quality: test policy\u002Fstrategy, environment record, execution\u002Fincident log, summary report, V&V report, quality gates, CAPA, assurance case;\n- operations\u002Fsupport: operations plan, service reports, incident\u002Fpostmortem, problem\u002Fknown-error records, service desk, recovery rehearsal;\n- maintenance\u002Fretirement: transition plan, data migration, archive, decommissioning;\n- security: security plan, risk assessment\u002Facceptance, incident response, posture and dependency-compliance reports;\n- process and organizational improvement: process tailoring, measurement repository, retrospective\u002Flessons-learned, improvement plan.\n\nUse “missing as a named category” when an existing broader row may contain the content. Do not claim a concept is wholly absent if the source material or target contains it under another name.\n\n### 5. Audit priorities using context, not intuition\n\nStart from the target’s own priority definition. For each Must Have, ask:\n\n- Is the underlying activity essential, or only the separate document name?\n- Is it applicable to virtually all projects, or only to a system type\u002Flifecycle\u002Fdomain?\n- Can it be represented by an SRS, backlog, repository, issue tracker, CI system, or dashboard?\n- Does the source explicitly require it, or merely mention it as an example?\n\nFlag **over-classification** when a conditional artifact is marked universal: formal architecture packages, ADRs\u002F4+1 views, SBOMs, UAT sign-offs, SLAs, DR plans, FCA\u002FPCA, QMS, formal V&V, and security reports are common examples. Flag **under-classification** when a source calls something a core activity\u002Foutput but the catalog marks it Optional\u002FNice, such as requirements baselining, status accounting under formal CM, or recovery evidence for an in-scope production service.\n\nPrefer a three-axis recommendation when possible:\n\n- **Core practice:** generally expected activity;\n- **Artifact form:** separate document, record, tool-managed item, or generated report;\n- **Applicability:** universal, common, conditional, or domain-specific.\n\n### 6. Detect duplicates by relationship, not just exact text\n\nClassify overlaps as:\n\n- **Exact duplicate:** same concept and scope under two headings;\n- **Parent\u002Fsubset:** e.g., test suite contains test cases; SAST is a security subset of static analysis;\n- **Lifecycle variant:** same artifact at requirements, design, implementation, or maintenance scope;\n- **Format\u002Fcontent duplicate:** ADR and design rationale, release notes and VDD, API contract and API reference docs;\n- **Process versus record:** incident-management process versus incident record\u002Fpostmortem;\n- **Cross-cutting view:** one end-to-end RTM versus phase-specific traceability views.\n\nRecommend a single canonical category with scope\u002Ftype fields instead of deleting useful distinctions.\n\n### 7. Verify source and standard attributions\n\nFor each citation, check four things:\n\n1. **Scope:** does the standard govern this artifact or only a neighboring process?\n2. **Edition\u002Fstatus:** is the cited edition current, withdrawn, superseded, or unspecified?\n3. **Type:** is it an international standard, IEEE standard, framework, specification, taxonomy, regulation, tool, or book?\n4. **Local support:** does the cited source file actually mention or support the attribution?\n\nCommon errors include using a quality model as a report template, an ISMS standard as a SAST\u002Fsecurity-architecture standard, an architecture-description standard as an ADR-format standard, an open-source compliance standard as an SBOM standard, or a generic UML citation without a version. Separate columns for `Source\u002FKA`, `Artifact standard`, `Framework\u002Ftaxonomy`, and `External practice` are preferable.\n\nCompletion criterion: every “incorrect” attribution is supported either by the authoritative standard’s stated scope or by a clear mismatch between the target citation and the local source text.\n\n### 8. Write the report, not a replacement catalog\n\nUse this compact structure:\n\n1. **Scope and headline finding** — roots audited, entry count, major pattern.\n2. **Missing categories** — grouped by lifecycle\u002Fknowledge area, with source and target references.\n3. **Priority errors** — over-classified, under-classified, and context-dependent items.\n4. **Duplicates\u002Foverlaps** — original rows plus recommended relationship.\n5. **Source\u002Fstandard corrections** — precise citation issues and evidence.\n6. **Files changed** — explicitly say none when the audit was read-only.\n7. **Tooling issue** — mention only if it materially affected the method; describe the workaround, not a permanent refusal.\n\nKeep conclusions proportional to evidence. Use “not named,” “overlaps,” “weakly supported,” or “conditional” where absolute claims would overstate the comparison.\n\n## Common Pitfalls\n\n1. **Treating SWEBOK as a mandatory document template.** SWEBOK is a body-of-knowledge map; distinguish generally accepted practice from project-specific work-product choices.\n2. **Counting a separate file as the essential thing.** The essential thing may be traceability, change control, testing evidence, or operational knowledge, represented in a tool or combined document.\n3. **Using absence as proof without checking aliases.** Search for synonyms such as vision\u002Fscope versus BRD, status report versus dashboard, postmortem versus RCA, and deployment package versus release artifact.\n4. **Calling every overlap a duplicate.** Preserve useful lifecycle or scope distinctions; report parent\u002Fsubset and process\u002Frecord relationships.\n5. **Applying a standard outside its scope.** Verify whether it defines a process, a data format, a quality model, an ISMS, a review framework, or an artifact template.\n6. **Ignoring conditional rows.** Reclassify by applicability instead of forcing one universal priority.\n7. **Citing a local summary as if it were the primary standard.** Identify whether the note is a paraphrase, a book-derived practice, or the actual authoritative publication.\n8. **Replacing the requested audit with edits.** The user asked for evidence; do not modify the catalog unless explicitly requested.\n9. **Overstating current-edition facts.** Include the edition\u002Fyear and status when the claim depends on currency.\n10. **Trusting prior remediation claims.** When the target set contains an earlier audit or review doc marked \"✅ FIXED\", re-verify every claimed fix with a machine check (grep\u002Fscript) before accepting it — quick passes routinely claim blanket fixes that only partially landed, and repeating the claim launders a false fix into the next baseline.\n11. **Patching a self-contradictory index.** When a master checklist's totals disagree with each other AND with the files on disk, do not renumber it in place; flag it for regeneration-from-disk after the categorization decision, because hand-maintained counts drift again immediately.\n\n## Obsidian\u002FMarkdown Template-Set Audits\n\nWhen the target is an Obsidian vault or Markdown repository, apply the additional probes below:\n\n- Treat the folder name `document_template` as a claim to test: distinguish a document catalog\u002Fchecklist from actual reusable templates with metadata and lifecycle fields.\n- Resolve Obsidian wikilinks and literal backtick paths separately; a healthy wikilink graph does not prove that embedded filesystem paths are current. Strip fenced code blocks before wikilink resolution — shell `[[ ... ]]` test operators otherwise register as broken links — and remember placeholder text with nested brackets (`[[X]%]`) also parses as a wikilink, a recurring false-broken source in placeholder-ized templates.\n- Compare profile counts and Quick-Start checklists against the priority-marked rows. Report both true omissions and abbreviation\u002Fnaming mismatches.\n- Build an applicability matrix that separates core practice, artifact form, and context (universal, common, conditional, domain-specific, or not applicable).\n- Reconcile overview\u002Findex counts and scope claims across the source corpus before treating extracted labels as authoritative.\n- Check repository status before\u002Fafter a read-only audit so pre-existing working-tree changes are not attributed to the audit.\n- When the templates feed AI-persona-driven spec generation, scan every template for example-content contamination (concrete IDs, currency, dates, named example entities that a consumer could inherit as real data) and report per-file scores, not just spot-checks.\n- When the user asks for a large restructure, the deliverable is a dated plan doc + companion data doc written into the vault — never in-place edits to the audited templates during the audit session.\n\nSee `references\u002Fobsidian-template-audit.md` for a reusable probe sequence, evidence patterns, standards-verification examples, and the contamination-scan scoring recipe.\n\n## Verification Checklist\n\n- [ ] Target file, source roots, and read-only scope are stated.\n- [ ] Target entries are counted and priority totals are grounded in inspection.\n- [ ] Every major finding has a target path\u002Fline and a source path\u002Fline where possible.\n- [ ] Missing categories distinguish absent names from absent concepts.\n- [ ] Priority findings distinguish core practice from standalone artifact and applicability.\n- [ ] Duplicate findings identify exact, parent\u002Fsubset, lifecycle, or process\u002Frecord relationships.\n- [ ] Standard attributions are checked for scope, edition\u002Fstatus, and type.\n- [ ] External authoritative links are included only for disputed\u002Fcurrent standard facts.\n- [ ] No files were edited or created unless the user explicitly authorized it.\n- [ ] The final report is concise enough to scan but specific enough to reproduce.\n\n## Supporting References\n\n- See `references\u002Fswebok-artifact-audit.md` for a condensed source map, artifact-category checklist, and standard-attribution traps from a completed SWEBOK audit.\n",{"name":5,"author":24,"license":15,"version":25,"metadata":26,"platforms":38,"description":7},"Hermes Agent","1.0.0",{"hermes":27},{"tags":28,"related_skills":35},[29,30,31,32,33,34],"research","document-audit","evidence","standards","swebok","traceability",[36,37],"privacy-regulatory-research","hermes-agent-skill-authoring",[39,40,41],"linux","macos","windows","main","2988869c5a7d7be2225c9ea7a54a4a56a59b0d45","workflow\u002Fproduct-owner\u002Fevidence-based-document-audit",0,"https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup","https:\u002F\u002Fgithub.com\u002Foat431","https:\u002F\u002Fpanomete.com\u002F","Hope and Luck is not a part of the strategy, \r\n\r\nSoftskill produces plan Hard skill produce product.","2026-11-03 00:43:08",{"source":8,"skillName":5,"commands":52,"prompt":62,"zipAvailable":63},[53,56,59],{"tool":54,"command":55},"npx","npx skills add oat431\u002Fhermes_config_backup --skill evidence-based-document-audit",{"tool":57,"command":58},"bunx","bunx skills add oat431\u002Fhermes_config_backup --skill evidence-based-document-audit",{"tool":60,"command":61},"pnpm","pnpm dlx skills add oat431\u002Fhermes_config_backup --skill evidence-based-document-audit","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fproduct-owner\u002Fevidence-based-document-audit\u002FSKILL.md",true,[65,71],{"path":66,"resourceType":67,"resourceTypeDesc":68,"sizeBytes":69,"sourceHtmlUrl":70},"references\u002Fobsidian-template-audit.md","REFERENCE","参考文档",10336,"https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fproduct-owner\u002Fevidence-based-document-audit\u002Freferences\u002Fobsidian-template-audit.md",{"path":72,"resourceType":67,"resourceTypeDesc":68,"sizeBytes":73,"sourceHtmlUrl":74},"references\u002Fswebok-artifact-audit.md",7471,"https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fproduct-owner\u002Fevidence-based-document-audit\u002Freferences\u002Fswebok-artifact-audit.md",[76,83,90,98,105,112],{"skillNo":77,"name":78,"nameZh":6,"description":79,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":80,"categories":81,"keywords":82,"publishedAt":19},"01M7AYZ2JY3NFKAPQSC39N5F3F","project-spec-docs","Fill\u002Fupdate project_spec docs (templates, MM handoffs).","https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fui-ux\u002Fproject-spec-docs\u002FSKILL.md",[],[],{"skillNo":84,"name":85,"nameZh":6,"description":86,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":87,"categories":88,"keywords":89,"publishedAt":19},"01M7AYYRZPE7HQX7778558X52S","penpot-mcp","Penpot MCP: connect, verify, and fix streaming via nginx.","https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fui-ux\u002Fpenpot-mcp\u002FSKILL.md",[],[],{"skillNo":91,"name":92,"nameZh":6,"description":93,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":94,"licenseFileUrl":6,"sourceHtmlUrl":95,"categories":96,"keywords":97,"publishedAt":19},"01M7AYYNKP61CN2281VMJDXMB5","nginx-streaming-proxy","Configure nginx for streaming endpoints (SSE, MCP, WS).","Unlicense","https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fui-ux\u002Fnginx-streaming-proxy\u002FSKILL.md",[],[],{"skillNo":99,"name":100,"nameZh":6,"description":101,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":102,"categories":103,"keywords":104,"publishedAt":19},"01M7AYYJP0MDV1D3WVEEZB3SGA","branding-ci","Use for brand\u002FCI identity, style guides, and brand docs.","https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fui-ux\u002Fbranding-ci\u002FSKILL.md",[],[],{"skillNo":106,"name":107,"nameZh":6,"description":108,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":94,"licenseFileUrl":6,"sourceHtmlUrl":109,"categories":110,"keywords":111,"publishedAt":19},"01M7AYYGCXVT97DDTCM0ZJT6F6","yuanbao","Yuanbao (元宝) groups: @mention users, query info\u002Fmembers.","https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fshared-skills\u002Fyuanbao\u002FSKILL.md",[],[],{"skillNo":113,"name":114,"nameZh":6,"description":115,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":94,"licenseFileUrl":6,"sourceHtmlUrl":116,"categories":117,"keywords":118,"publishedAt":19},"01M7AYYC2Y65NMGMD4GQXAZMMS","slidev","Create and present web-based slidedecks for developers using Slidev with Markdown, Vue components, code highlighting, animations, and interactive features. Use when building technical presentations, conference talks, code walkthroughs, teaching materials, or developer decks.","https:\u002F\u002Fgithub.com\u002Foat431\u002Fhermes_config_backup\u002Fblob\u002Fmain\u002Fworkflow\u002Fshared-skills\u002Fslidev\u002FSKILL.md",[],[]]