The Agent Skills Directory

[PROMPT_INJECTION]: The skill operates by reading and summarizing information from external sources such as source code, documentation, and changelogs. This architectural pattern introduces a vulnerability to indirect prompt injection, where an attacker could embed malicious instructions within these files to manipulate the agent's behavior during content generation.
Ingestion points: The skill utilizes tools to access local project files, including source code, documentation, and changelogs, as well as the internal references/style-guide.md file.
Boundary markers: No specific delimiters or safety instructions are present to prevent the agent from inadvertently following instructions found within the ingested content.
Capability inventory: The skill is restricted to file read operations and generating textual content for user approval, lacking high-risk capabilities like network access or arbitrary command execution.
Sanitization: The skill does not define any sanitization or filtering logic for the data retrieved from external files.

post-forge-feature