The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the analysis and modification of external, untrusted project files, which creates a surface for indirect prompt injection if the source code contains adversarial instructions.
Ingestion points: The 'Workflow' section in SKILL.md and the 'Inventory the feature' section in riverpod_migration.md instruct the agent to read pubspec.yaml, lockfiles, imports, and source code files.
Boundary markers: There are no specific instructions or delimiters defined in the skill to isolate external code content or to warn the agent to ignore instructions embedded within the data it processes.
Capability inventory: The skill grants the agent the ability to modify project files, format code, and execute local development tools such as dart analyze and dart test (as seen in SKILL.md and testing.md).
Sanitization: No sanitization, escaping, or validation logic is prescribed for the external data before it enters the agent's reasoning context.

bloc-signals