The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill utilizes the !command syntax in SKILL.md to execute several shell scripts (e.g., ops-infra, ops-prs) and a bash loop at load time. These commands replace placeholders with their output before the model processes the skill content.
[INDIRECT_PROMPT_INJECTION]: The skill ingests potentially untrusted data from Slack, Linear, and project-specific planning files without using explicit boundary markers or sanitization, creating a surface for indirect instructions.
Ingestion points: Data is pulled from Slack searches, Linear issue lists, and local .planning/STATE.md files.
Boundary markers: The instructions lack delimiters or warnings to ignore instructions embedded in the retrieved data.
Capability inventory: The skill has access to Bash, WebFetch, Agent (for spawning sub-agents), and task management tools.
Sanitization: There is no evidence of filtering or escaping external content before it is processed by the agent.
[COMMAND_EXECUTION]: The skill uses the Bash tool to run commands such as gh pr list, jq, and cat to extract information from the environment and configuration files.

ops-next