The Agent Skills Directory

[NO_CODE]: The skill consists entirely of Markdown instructions (SKILL.md) and a reference file (references/punctuation-guide.md). No executable scripts, binaries, or automated installation processes are included.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-supplied text for rewriting. This creates a surface where an attacker could embed hidden instructions within the source text. However, the skill does not use tools that permit network access, file system modifications, or command execution, which significantly limits the potential impact of such an injection.
Ingestion points: User-provided text for rewriting as described in SKILL.md.
Boundary markers: The instructions explicitly direct the agent to "Mark protected content such as code, commands, URLs, email addresses, citations, product names, identifiers, numbers, and quoted text" before rewriting, which serves as a defensive measure.
Capability inventory: No file writing, network requests, or subprocess execution capabilities are requested or used.
Sanitization: The skill emphasizes preserving meaning and facts, which acts as a logical check against manipulative injections.
[SAFE]: The instructions include specific safety constraints, such as prohibiting the agent from claiming to bypass AI detectors, concealing authorship, or fabricating personal experiences.

humanizer