The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from repository files to drive release logic and configure sub-agents, which could be exploited if these files are maliciously modified.
Ingestion points: Reads versioning and configuration data from production/session-state/active.md, production/milestones/, and production/review-mode.txt (referenced in references/claude-skill.md).
Boundary markers: There are no explicit delimiters or safety instructions provided when passing this file content to sub-agents.
Capability inventory: The skill utilizes powerful tools including Write, Edit, Bash, and Task to modify the codebase, execute shell commands, and spawn sub-processes.
Sanitization: The instructions do not include steps to sanitize or validate the content of the ingested files before processing.
[COMMAND_EXECUTION]: The skill is designed to execute shell commands via the Bash tool for deployment and release activities.
Evidence: Phase 6 in references/claude-skill.md instructs the agent to tag the release, generate changelogs, and deploy to staging and production environments.

team-release