The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses Bash, yq, git, and the gh CLI to automate product research workflows. It discovers configuration by searching parent directories for pulse-config.yaml and constructs Git commands using values parsed from this file. Potential command injection exists if the configuration file contains shell metacharacters, though the skill operates on the assumption of a trusted local research environment.
[INDIRECT_PROMPT_INJECTION]: The skill processes external research data and analyst briefs, creating a surface for indirect prompt injection. Ingestion points: Reads pulse-config.yaml, research-context.md, and historical reports like *-daily-research.md in SKILL.md. It also processes data from sub-agents performing web research. Boundary markers: Employs a structured delegation model via harness:delegate with specific worker constraints. Capability inventory: Possesses file-writing capabilities, Git repository modification, and PR management. Sanitization: Implements 'verification seams' where the agent must re-verify all claims against primary sources and includes a formal adjudication phase for resolving conflicting evidence.

weekly-strategist