Security audit
Security checks for vulnerabilities and agentic risk
This is a coherent Obsidian knowledge-wiki skill that reads user-selected or vault content and persistently updates a _wiki folder, with no evidence of hidden exfiltration or malicious behavior.
Install only in a vault where you are comfortable with the agent reading relevant notes and changing files under _wiki/. Use version control or backups, review generated wiki edits, and be careful with sensitive notes or untrusted URLs because the skill intentionally synthesizes and persists model-generated content.
The README promotes ingesting from 'a file, URL, or topic' without any privacy or data-handling notice about sending external content and potentially local vault data to an LLM-backed workflow. Users may unknowingly process sensitive notes or fetch untrusted remote content, creating confidentiality and provenance risks.
The README explicitly says query operations 'file new insights back' and that the wiki compounds over time, but it does not clearly warn users that using the skill will modify files in their Obsidian vault. For a tool operating on user knowledge bases, silent or poorly disclosed write behavior can cause unintended data alteration, trust issues, and accidental propagation of incorrect model-generated content.
Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.
| Sub-command | Delegates to | Description |
|---|---|---|
| `ingest <source>` | `skills/llmwiki-ingest/SKILL.md` | Extract insights from a file, URL, or topic into `_wiki/` |
| `query <question>` | `skills/llmwiki-query/SKILL.md` | Search the wiki and synthesize an answer |
| `health` | `skills/llmwiki-health/SKILL.md` | Audit wiki for orphans, stubs, broken links, stale pages |
Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.
| Sub-command | Delegates to | Description |
|---|---|---|
| `ingest <source>` | `skills/llmwiki-ingest/SKILL.md` | Extract insights from a file, URL, or topic into `_wiki/` |
| `query <question>` | `skills/llmwiki-query/SKILL.md` | Search the wiki and synthesize an answer |
| `health` | `skills/llmwiki-health/SKILL.md` | Audit wiki for orphans, stubs, broken links, stale pages |
Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.
| Sub-command | Delegates to | Description |
|---|---|---|
| `ingest <source>` | `skills/llmwiki-ingest/SKILL.md` | Extract insights from a file, URL, or topic into `_wiki/` |
| `query <question>` | `skills/llmwiki-query/SKILL.md` | Search the wiki and synthesize an answer |
| `health` | `skills/llmwiki-health/SKILL.md` | Audit wiki for orphans, stubs, broken links, stale pages |
Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.
| Sub-command | Delegates to | Description |
|---|---|---|
| `ingest <source>` | `skills/llmwiki-ingest/SKILL.md` | Extract insights from a file, URL, or topic into `_wiki/` |
| `query <question>` | `skills/llmwiki-query/SKILL.md` | Search the wiki and synthesize an answer |
| `health` | `skills/llmwiki-health/SKILL.md` | Audit wiki for orphans, stubs, broken links, stale pages |
If no sub-command is given, print this help:
Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.
| Sub-command | Delegates to | Description |
|---|---|---|
| `ingest <source>` | `skills/llmwiki-ingest/SKILL.md` | Extract insights from a file, URL, or topic into `_wiki/` |
| `query <question>` | `skills/llmwiki-query/SKILL.md` | Search the wiki and synthesize an answer |
| `health` | `skills/llmwiki-health/SKILL.md` | Audit wiki for orphans, stubs, broken links, stale pages |
If no sub-command is given, print this help:
Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.
| Sub-command | Delegates to | Description |
|---|---|---|
| `ingest <source>` | `skills/llmwiki-ingest/SKILL.md` | Extract insights from a file, URL, or topic into `_wiki/` |
| `query <question>` | `skills/llmwiki-query/SKILL.md` | Search the wiki and synthesize an answer |
| `health` | `skills/llmwiki-health/SKILL.md` | Audit wiki for orphans, stubs, broken links, stale pages |
If no sub-command is given, print this help:
Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.
|---|---|---|
| `ingest <source>` | `skills/llmwiki-ingest/SKILL.md` | Extract insights from a file, URL, or topic into `_wiki/` |
| `query <question>` | `skills/llmwiki-query/SKILL.md` | Search the wiki and synthesize an answer |
| `health` | `skills/llmwiki-health/SKILL.md` | Audit wiki for orphans, stubs, broken links, stale pages |
If no sub-command is given, print this help:
Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.
|---|---|---|
| `ingest <source>` | `skills/llmwiki-ingest/SKILL.md` | Extract insights from a file, URL, or topic into `_wiki/` |
| `query <question>` | `skills/llmwiki-query/SKILL.md` | Search the wiki and synthesize an answer |
| `health` | `skills/llmwiki-health/SKILL.md` | Audit wiki for orphans, stubs, broken links, stale pages |
If no sub-command is given, print this help:
Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.
|---|---|---|
| `ingest <source>` | `skills/llmwiki-ingest/SKILL.md` | Extract insights from a file, URL, or topic into `_wiki/` |
| `query <question>` | `skills/llmwiki-query/SKILL.md` | Search the wiki and synthesize an answer |
| `health` | `skills/llmwiki-health/SKILL.md` | Audit wiki for orphans, stubs, broken links, stale pages |
If no sub-command is given, print this help:
The skill explicitly states it will 'fix what can be fixed automatically' and later instructs the agent to modify wiki files, but it does not warn the user that invoking the command may change repository content. In an LLM-agent setting, silent file edits are risky because users may expect an audit/report-only action and instead receive unintended content changes that alter knowledge integrity or introduce bad links/metadata.
The auto-fix section directs the agent to add wikilinks and fill frontmatter fields from page content, which are content-altering operations performed without any approval gate or safety constraints. Because these edits are heuristic and LLM-driven, they can create incorrect relationships, fabricate metadata, or overwrite user intent, reducing trustworthiness of the wiki and causing cascading quality issues.
The skill instructs the agent to create or update _wiki/ pages and append entries to _wiki/log.md, but it does not clearly warn the user that invoking the skill will modify vault contents. This can lead to unintended file changes, unwanted growth of logs or stub pages, and loss of user trust because the skill performs persistent writes as part of normal operation.
The skill description promises that queries will automatically write new insights back into _wiki/, but it does not warn the user that invoking a read-like query action may modify stored files. This creates an integrity risk because the model can persist hallucinated, speculative, or prompt-influenced content into the knowledge base without explicit user consent.
The workflow explicitly instructs the agent to update or create wiki pages whenever it derives something new, yet it provides no user-facing warning or approval gate for modifying persistent knowledge files. In this context, a query skill is expected to be primarily read-oriented, so silent writes increase the chance of poisoning the wiki with inaccurate summaries, adversarial content from searched notes, or unintended changes to curated knowledge.
No suspicious patterns detected.