The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill reads and processes the CHANGELOG.md file to automate version heading updates and link generation.
Ingestion points: Content is read from CHANGELOG.md using filesystem tools.
Boundary markers: The skill lacks explicit markers or instructions to delimit untrusted file content from the agent's instructions.
Capability inventory: The agent utilizes run_terminal_cmd to execute git operations.
Sanitization: The skill applies semantic version validation and follows a specific replacement logic, which provides a basic level of protection against malicious formatting.
[PROMPT_INJECTION]: The instructions refer to an external file named AGENTS.md for Operational Standards and Boundaries. Referencing local files for behavioral guidelines is a pattern that could be exploited to inject instructions or override agent safety parameters if the file content is modified by an attacker.
[COMMAND_EXECUTION]: The skill generates shell commands for git operations such as tag and commit. It uses a regex-based validation step (v[0-9]+\.[0-9]+\.[0-9]+) for the version input, which helps mitigate basic command injection attempts via the version parameter.

create-release