The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is instructed to read file contents and git diffs to understand the intent of changes. This represents an ingestion point for untrusted data that could contain malicious instructions designed to influence the agent's behavior during the commit grouping or validation phases.
Ingestion points: Git diffs and file reads (SKILL.md, Workflow step 2).
Boundary markers: Absent; no delimiters or instructions to ignore embedded commands are specified for the ingested content.
Capability inventory: Shell command execution (git), network operations (git push), and execution of repository quality gates.
Sanitization: Absent; the skill does not specify filtering or validation of the content read from the repository.
[COMMAND_EXECUTION]: The skill uses the system git CLI to perform status checks, diffing, staging, committing, and pushing. It also executes 'quality gates' based on the repository's own configuration (e.g., test scripts or build commands).
[DATA_EXFILTRATION]: The skill's primary workflow involves pushing local repository data to remote servers (git push). While this is the intended purpose, it facilitates the movement of local code and potential configuration data to external systems.
[DYNAMIC_EXECUTION]: The skill is instructed to run 'quality gates' using the repository's own checks. This involves discovering and executing arbitrary scripts or commands defined within the project environment (SKILL.md, Workflow step 4).

smart-commits