The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses dynamic context injection (the ! command syntax) to execute ${CLAUDE_PLUGIN_ROOT}/scripts/discover-config.sh at load time. This pattern executes shell commands automatically when the skill is accessed, replacing placeholders with script output.
[COMMAND_EXECUTION]: In Phase 3.1, the skill uses eval to execute the shell output generated by ${CLAUDE_PLUGIN_ROOT}/scripts/for-each-board.sh. This pattern is risky as it executes dynamically generated code, making the execution flow dependent on the integrity of the external script and its input parameters.
[COMMAND_EXECUTION]: The skill performs active repository management, including git checkout and git pull operations on configured paths. This involves network activity and local filesystem modifications based on configuration variables.
[PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection. It ingests untrusted data (research reports) from research directories and passes the content to a sub-agent (ingestion-analyst.md).
Ingestion points: Research reports found in directories specified by the research_dirs configuration.
Boundary markers: None explicitly defined in the provided file to separate report content from agent instructions.
Capability inventory: The skill has the capability to write to the local filesystem (YAML items), execute shell commands (gh issue create, git pull), and interact with the Trello API via MCP tools.
Sanitization: There is no evidence of sanitization or validation of the content extracted from research reports before it is processed by the agent or used to populate issue tracker fields.

ingest