The Agent Skills Directory

[PROMPT_INJECTION]: The skill provides documentation for patterns that inges t and proces s untrusted data from external sources, whic h present s an in direct prompt injection surface.
Ingestion points: External data is processed via FastAPI end points (on_input in references/pattern-debouncing.md), webhooks (payment_webhook in references/com m-messages.md), and workflow arguments (prompt in references/com m-streamin g.md).
Boundary markers: The examples do no t include explici t boundary markers or instructions to isolat e external data from the workflow execution logic.
Capabil ity inventory: Documented capabil ities include database acces s via DBOS.sql_session.execute and network operations usin g requests.get and aiohttp.
Sanitization: No specifi c inpu t sanitization or validation routines are show n in the data-handlin g snippets, althoug h SQL parameterization is correctl y used.
[SAFE]: The skill recommends usin g environment variables (os.environ) for managin g database connection strings and othe r sensitive configurations, followin g standar d secre t managemen t practices.
[SAFE]: The provide d database transaction pattern s use SQLAlchemy's parameterize d queries, effectivel y preventin g SQL injection vulnerabil ities.

dbos-python