The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The Tier 2 discovery protocol generates a repository outline consisting of pubspec.yaml dependencies, directory structures, and API signatures which are then passed to an LLM. This process creates a surface where malicious or deceptive content within the analyzed codebase could attempt to influence the agent's architectural assessment.
Ingestion points: File system structure, pubspec.yaml, and Dart source code signatures (SKILL.md).
Boundary markers: The instructions do not specify the use of delimiters or instructions to ignore embedded content when passing the outline to the model.
Capability inventory: The skill executes dart run dash_discover and sem entities to extract architectural data (SKILL.md).
Sanitization: There is no mention of sanitizing or escaping the extracted signatures before they are interpolated into the LLM prompt.
[EXTERNAL_DOWNLOADS]: The remediation workflow encourages the agent to follow github_url links found in the tool's output to inspect or install recommended specialized skills. This involves fetching code from dynamic external sources based on the analysis of the local repository.
[COMMAND_EXECUTION]: The skill relies on the execution of CLI tools such as dart run dash_discover and sem entities to perform its primary function of repository scanning and outline generation.

dash-discover