The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources that could potentially contain malicious instructions.\n
Ingestion points: Feature implementation tasks are ingested from local markdown and YAML files in the scripts/implementation_manager.py script. Additionally, sub-issue data is fetched from GitHub issue bodies, and agent transcripts are read in scripts/task_status_hook.py.\n
Boundary markers: The instructions do not specify any boundary markers or delimiters to differentiate between task metadata and potentially malicious natural language instructions embedded in task descriptions.\n
Capability inventory: The skill possesses the capability to execute shell commands via subprocess for task management tools (sam_cli, bd) and perform local file system operations.\n
Sanitization: Although task data is parsed into structured formats (JSON/YAML) and some IDs are validated with regular expressions, the natural language content of the tasks is not sanitized before being processed by the agent.\n- [COMMAND_EXECUTION]: The skill executes authorized local CLI tools and internal scripts to query and update implementation states. These operations are limited to specific management tasks and use structured arguments.\n- [EXTERNAL_DOWNLOADS]: The skill connects to GitHub to retrieve task backlog information. This communication is performed for the intended purpose of the skill and targets a well-known service provider.

implementation-manager