The Agent Skills Directory

[DATA_EXFILTRATION]: The skill is configured to read and parse sensitive configuration files, specifically .env files, which are commonly used to store environment secrets and credentials. While the stated purpose is to identify architectural dependency hints (like database or message queue names), the tool performs full file reads on these sensitive paths.
Evidence: The logic in src/architecture_path_extractor/extractors/config_parser.py explicitly allows files ending in .env to be read and searched for keywords.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted data from external repositories and documents to generate its output, without implementing sanitization or boundary markers to prevent embedded instructions from influencing the agent's behavior.
Ingestion points: The scan_repo_files function in src/architecture_path_extractor/extractors/repo_scanner.py and various parsers in the extractors/ directory ingest arbitrary file content from the targeted repository.
Boundary markers: None identified in the processing pipeline or output generation.
Capability inventory: The skill has the capability to write multiple files to the local file system via src/architecture_path_extractor/exporter.py.
Sanitization: No sanitization or filtering of external content is performed before processing or inclusion in output reports.

byted-sol-stability-architecture-path-extractor