The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it is designed to ingest and analyze untrusted source code and dependency files. Adversarial instructions embedded in these files could potentially influence or override the agent's behavior during the analysis phase.
Ingestion points: Target source code files and their dependencies identified via import analysis as specified in SKILL.md and rules/general/code-context-analysis.md.
Boundary markers: Absent. The instructions do not implement delimiters or provide explicit warnings to the agent to ignore instructions found within the analyzed content.
Capability inventory: The agent uses Read, Glob, and Grep to access and process the content of external files.
Sanitization: Absent. No filtering or content validation is performed on the data read from the filesystem.
[DATA_EXFILTRATION]: The skill presents a risk of data exposure by instructing the agent to read arbitrary file paths provided via the $ARGUMENTS variable without enforcing path restrictions or validation. This allows the agent to be directed to read and summarize sensitive files on the system (such as environment variables or credentials) if requested by a user or an adversarial prompt.
Evidence: The instructions in SKILL.md direct the agent to "Read the target source file/class/method specified above" based on user input, providing no mechanism to restrict the file access to appropriate project directories.

generate-test-cases