The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on the pythia CLI utility to interact with the database environment. It executes subcommands including errors for compilation status, src for source retrieval, impact for dependency analysis, and similar for style consistency checks.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes PL/SQL source code from the database, which is an untrusted ingestion point.
Ingestion points: Database source code is loaded into the agent's context via the pythia src command in SKILL.md.
Boundary markers: Absent. There are no instructions to wrap the ingested code in protective delimiters or to treat it as data only.
Capability inventory: The agent can execute database introspection commands through the pythia tool.
Sanitization: No sanitization or filtering logic is defined for the ingested source code content.

plsql-review