The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process and analyze untrusted data sources including application logs, API responses, database records, and browser console output. This introduces a vulnerability surface where malicious instructions hidden in logs or responses could attempt to influence the agent's behavior during a debugging session.
Ingestion points: The agent ingests data through commands like docker compose logs, curl, and psql (SKILL.md).
Boundary markers: There are no explicit markers or instructions provided to the agent to treat this ingested content as untrusted or to ignore embedded instructions.
Capability inventory: The skill utilizes powerful capabilities including shell command execution inside containers (docker compose exec), network requests (curl), and file modification for bug fixing.
Sanitization: No sanitization or validation of the ingested external content is mentioned.
[CREDENTIALS_UNSAFE]: The skill includes a hardcoded credential (secret) in a PostgreSQL connection string example (postgresql://app:secret@localhost:5432/app). While intended for local development, providing credentials in instructions is a risky practice that could lead to accidental exposure if misused in other environments.
[COMMAND_EXECUTION]: The skill relies on extensive use of shell commands to interact with the system and containers. This includes docker compose exec for interactive shell access and various monitoring tools, which are powerful capabilities that require careful oversight.

debug