The Agent Skills Directory

[PROMPT_INJECTION] (LOW): Indirect Prompt Injection risk through ingestion of untrusted external data.
Ingestion points: The skill reads GitHub PR descriptions and linked issues via the gh cli tool.
Boundary markers: There are no specified delimiters or warnings to ignore instructions embedded in the PR data.
Capability inventory: The skill uses gh cli for reading and potentially curl for network output to Discord.
Sanitization: No sanitization or escaping of the ingested PR content is performed before generating the report.
[DATA_EXFILTRATION] (LOW): The skill instructions explicitly encourage the inclusion of potentially sensitive deployment data in the final output.
Evidence: The "Deployment Notes" section asks the agent to include "Database migrations required" and "Environment variable updates needed".
Risk: If these notes contain actual secrets or sensitive schema details, they are sent to an external Discord webhook.
[COMMAND_EXECUTION] (LOW): The skill provides a shell script template using curl to post data to an external Discord webhook.
Evidence: The "Discord Posting" section contains a code block with a curl command that interpolates the {{CHANGELOG}} variable.
Risk: If the agent executes this command using a shell, an attacker-controlled PR description could attempt to break out of the JSON payload and execute arbitrary commands or leak environment variables via shell injection.

changelog