Security audit
Security checks for vulnerabilities and agentic risk
This is a disclosed writing-audit and rewrite skill whose file-editing behavior is user-directed and proportionate to its purpose.
Install this if you want an assistant to audit or rewrite prose for AI-like wording. Review edits before accepting them, especially in edit-in-place mode, and avoid using its AI-writing signals as proof of authorship or for consequential judgments.
This markdown file explicitly documents invocation phrases, so SQP-1 applies. The trigger list includes broad terms such as "scan" and the open-ended "or similar," without constraints or negative examples, making activation scope ambiguous and increasing the chance of unintended invocation.
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.
# avoid-ai-writing
Audit & rewrite content to remove AI writing patterns. A practical skill for any AI agent. Supports detect-only and edit-in-place modes, plus voice profiles.
[](https://github.com/conorbronsdon/avoid-ai-writing/stargazers)
[](LICENSE)
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.
$ARGUMENTS
Read and follow the instructions in ~/.claude/skills/avoid-ai-writing/SKILL.md
Then use `/clean-ai-writing <your text>` in Claude Code.
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.
$ARGUMENTS
Read and follow the instructions in ~/.claude/skills/avoid-ai-writing/SKILL.md
Then use `/clean-ai-writing <your text>` in Claude Code.
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.
Cowork loads skills only from installed plugins — it doesn't scan ~/.claude/skills/, so a bare clone (the Claude Code steps above) won't be discovered there. This repo doubles as a single-plugin marketplace, so install it as a plugin instead:
/plugin marketplace add conorbronsdon/avoid-ai-writing
The README says the skill auto-triggers from phrases like "remove AI-isms," but does not define a bounded trigger list, exclusions, or contextual constraints for when auto-triggering should or should not happen. In a markdown skill description, this kind of open-ended phrasing can cause unintended invocation from ordinary editing requests that mention AI writing in passing.
The trigger guidance includes generic terms such as "scan" and the catch-all "or similar," which broadens activation beyond clearly defined requests. Because this is a markdown file, ambiguous invocation wording is in scope and may lead to accidental detect-mode activation during unrelated requests.
This markdown file defines invocation behavior, so vague-trigger review applies. The detect-mode trigger list ends with "or similar," which leaves the activation boundary open-ended and makes it unclear what nearby phrasings should or should not invoke the skill.
The invocation guidance allows free-form natural language without clearly constraining what requests count as invocation versus ordinary discussion about AI writing. That broad wording increases the risk of unintended activation collisions in normal conversation.
No suspicious patterns detected.