The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes commands discovered within repository configuration files (such as package.json scripts or instructions) using the herdr pane run command.
Evidence: Found in references/operations.md under the 'Optional runtime pane' section: herdr pane run "$runtime_pane" "$verified_repo_command".
Mitigation: The skill includes explicit instructions in SKILL.md (Section 2) and references/operations.md to present the plan to the user and obtain approval before applying any mutations or running commands.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources including project handoffs, notes vaults, and repository instructions to configure its operating environment and seed agents with context.
Ingestion points: SKILL.md mentions reading from a 'notes vault' and 'selected checkouts'; references/operations.md mentions 'package scripts'.
Boundary markers: The skill explicitly defines a 'no-write boundary' for external context and instructs agents to verify claims before mutation.
Capability inventory: The skill can execute shell commands via herdr and launch independent AI agents.
Sanitization: The workflow relies on human verification of the 'mutation plan' before any resources are created or scripts are executed.

herdr-workstreams