The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources including GitHub Issues, Asana tasks, and Jira tickets. This creates a potential surface for indirect prompt injection. However, the skill explicitly mitigates this through a 'Third-Party Content Safety' section which mandates that fetched content be treated as data only, requires a user confirmation gate before processing, and instructs the agent to flag and ignore common injection patterns.
[PROMPT_INJECTION]: A static detector flagged strings related to instruction overrides. Analysis confirms these are part of a defensive pattern-matching guide for the agent to identify malicious input in external data, rather than an attempt to compromise the agent itself.
[COMMAND_EXECUTION]: The skill uses the GitHub CLI (gh) and Asana MCP tools to fetch and manage requirements. These operations are restricted to relevant fields and are well-aligned with the skill's primary function of requirements management.

usdm