The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/findings_to_plan.py executes the local shell script scripts/redundancy_scan.sh using the subprocess.run function. The execution uses a list-based argument structure which prevents shell injection. The script is used to search the local codebase using ripgrep for structural patterns.
[DATA_EXFILTRATION]: The skill performs repository-wide searches and reads code files to generate its analysis. This data processing is entirely local, and there are no signs of network requests, external data transfers, or hardcoded credentials.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted repository content during its analysis phase.
Ingestion points: scripts/redundancy_scan.sh (via rg) and scripts/findings_to_plan.py (via file reading and output parsing).
Boundary markers: The skill does not implement specific delimiters or 'ignore' instructions when processing the names of symbols or snippets of code from the files.
Capability inventory: The skill has the ability to execute local shell scripts via subprocess.run and write plan files to the local disk using Path.write_text.
Sanitization: The skill uses regular expressions to extract specific architectural symbols (structs, enums, traits) rather than interpreting the entire file content as instructions, which reduces the potential impact of malicious payloads in the analyzed source code.

plan-flow