The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted data from external sources, which could contain malicious instructions designed to subvert agent behavior.
Ingestion points: Step 1 in SKILL.md instructs the agent to fetch external URLs, read user-provided files, and use web search to gather information.
Boundary markers: The skill lacks explicit delimiters or instructions to ignore potential commands embedded within the summarized resource files.
Capability inventory: The skill has the ability to write files, perform web searches, and execute git commands (commit, push).
Sanitization: No evidence of sanitization, filtering, or validation of the content retrieved from external sources before it is used to generate the outline and draft.
[COMMAND_EXECUTION]: The skill executes shell commands to manage the file system and version control.
Evidence: SKILL.md contains instructions for folder creation (Step 0) and Git operations including git commit and staging files (Steps 5 and 7).
[DATA_EXFILTRATION]: While intended for version control, the skill performs automated network operations that transmit local content to external servers.
Evidence: Steps 5 and 7 in SKILL.md instruct the agent to execute git push to a remote repository after committing blog post drafts and resources.

blog-draft