The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the Beads issue tracking system, creating a surface for indirect prompt injection. * Ingestion points: Data is collected via bd list and bd dep commands as defined in data/bd-commands.md. * Boundary markers: The instructions lack explicit boundary markers or delimiters to distinguish issue data from the agent's core instructions. * Capability inventory: The agent uses the bd CLI for data retrieval and is instructed to 'PROACTIVELY create OpenSpec proposal drafts', implying file system write capabilities. * Sanitization: No sanitization or filtering of issue content is described in the analysis process.
[COMMAND_EXECUTION]: The skill relies on executing the bd CLI tool to perform its primary functions. While the commands listed in data/bd-commands.md are scoped to listing and dependency mapping, the agent's interaction with this external tool is a core component of its operational logic.

beads-retrospective