The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's schema inspection workflow creates a surface for indirect prompt injection by ingesting metadata from live databases.
Ingestion points: The scripts/pull-hominem-schema.sh script queries information_schema.tables and information_schema.columns to generate schema snapshots.
Boundary markers: The generated snapshots lack explicit delimiters or instructions to the agent to ignore potentially malicious content within table or column names.
Capability inventory: The skill context includes capabilities for executing shell commands (psql, make, just, pnpm) and modifying the file system (writing snapshots).
Sanitization: Data retrieved from the database is not sanitized or escaped before being presented to the agent in the snapshot files.
[CREDENTIALS_UNSAFE]: Default connection strings containing credentials for a local development PostgreSQL instance are hardcoded in scripts and documentation.
Evidence: scripts/pull-hominem-schema.sh and references/hominem-workflow.md contain the string postgresql://postgres:postgres@127.0.0.1:5434/hominem. While suitable for local development environments, these represent hardcoded credentials.

kernel-dev-database