The Agent Skills Directory

[COMMAND_EXECUTION]: The skill implements a custom rule-based workflow where execution logic is defined as embedded Python snippets within YAML files (e.g., rules/evidence_first_rules.yaml, rules/task_freeze_rules.yaml). These snippets perform operations such as state transitions, file writes, and subagent spawning at runtime.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests and processes task data that can be influenced by external actors (e.g., via GitHub issue descriptions).
Ingestion points: The skill reads task.description and task.blocker_details within the action blocks of rules/model_routing_rules.yaml and rules/task_freeze_rules.yaml.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the rule evaluation logic.
Capability inventory: The skill utilizes spawn_subagent, write_to_file, append_to_file, and save_to_file commands across its YAML rule definitions to manage task state and memory.
Sanitization: No sanitization or filtering of external input is observed before it is used in conditional logic or logged to memory files.

bounty-hunter-starter