The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external websites to perform its analysis, creating an attack surface for indirect prompt injection.
Ingestion points: The skill fetches visible text, CSS styles, meta tags, and font imports from user-provided URLs using the WebFetch tool, as described in Step 2 of SKILL.md.
Boundary markers: There are no explicit instructions or delimiters provided to help the agent distinguish between the website's data and the skill's instructions.
Capability inventory: The skill utilizes WebFetch for network reading and writes the resulting analysis to the local file system as brand-profile.json.
Sanitization: While the skill instructs the agent to use null for non-extractable fields, there is no explicit sanitization or filtering to prevent the model from following instructions found within the fetched content.
[EXTERNAL_DOWNLOADS]: The documentation provides manual installation instructions that reference a remote repository.
Evidence: The Install section in SKILL.md suggests cloning the repository https://github.com/thatrebeccarae/claude-marketing.git.
The repository is owned by the skill's author.

brand-dna