The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted data from the repository during the validation and release cycle.
Ingestion points: Documentation files in docs/user/, Conventional Commit messages, and RELEASING.md in SKILL.md.
Boundary markers: No explicit delimiters or warnings to ignore embedded instructions are present in the documentation validation steps.
Capability inventory: The agent is authorized to execute npm commands, read multiple local files, and execute locally built binary artifacts.
Sanitization: There is no evidence of sanitization or filtering of the content read from the project documentation or commit history.
[DYNAMIC_EXECUTION]: The workflow involves generating and then running executable code as part of the release validation.
Execution pattern: The agent is instructed to run npm run build:executable and then execute the resulting binary located in ./artifacts/.
Risk factor: While this is a standard local build-and-test procedure, the execution of dynamically created binaries represents a high-capability interaction with the host system.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill mentions the use of sensitive authentication tokens for publishing and CI/CD.
Evidence: Reference to NPM_TOKEN and RELEASE_PLEASE_TOKEN in the Diagnosticar falhas section of SKILL.md.
Context: Although these are intended to be managed as environment secrets, the agent's role in diagnosing issues related to these tokens could lead to accidental exposure if the agent is prompted to reveal environment variables or configuration details.

clickupfy-release