The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (SAFE): No external dependencies, packages, or remote scripts are downloaded. All validation logic and rules are contained within local Bash scripts and JSON/Markdown templates.
[COMMAND_EXECUTION] (SAFE): The skill executes local validation scripts and standard Linux utilities (grep, find, sed, awk) to parse SQL files. It optionally attempts to run psql --dry-run for syntax verification if the PostgreSQL client is available on the host, which is a legitimate use case for schema validation.
[DATA_EXFILTRATION] (SAFE): No network operations (curl, wget, fetch) were detected. The skill only performs local file read/write operations to generate a validation report.
[PROMPT_INJECTION] (SAFE): No instructions targeting agent behavior, safety filter bypass, or role-play injections were found in the markdown files or metadata.
[CREDENTIALS_UNSAFE] (SAFE): No hardcoded API keys, tokens, or database credentials were found. The skill operates on schema files and does not require active database connections beyond an optional local syntax check.
[INDIRECT_PROMPT_INJECTION] (LOW): As a static analysis tool, the skill reads user-provided SQL files. While this creates a surface for indirect prompt injection if an attacker embeds instructions in SQL comments that the agent later reads in the generated report, this is a standard risk for development tools and is mitigated by the structured nature of the report output.

schema-validation