The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external data sources which may contain untrusted content.
Ingestion points: The workflow reads retrospective notes from production/sprints/, velocity data from production/session-state/, commit history via git log, and issue content via gh issue list.
Boundary markers: The instructions do not define specific delimiters or instructions for the agent to ignore potentially malicious content embedded within these external files or command outputs.
Capability inventory: The skill utilizes Bash, Write, Edit, and Read tools, providing a surface for actions if the agent is influenced by injected instructions.
Sanitization: There are no explicit instructions to sanitize, escape, or validate the content of retrospective notes or issue descriptions before they are processed by the LLM during Phase 3 (Pattern Analysis).
[DYNAMIC_CONTEXT_INJECTION]: The skill uses the !command syntax to automatically populate project metadata when the skill is loaded.
Evidence: Multiple instances in the 'Project Data (Auto-populated)' section, including !git rev-list --count HEAD, !git log, !gh issue list, and !ls production/sprints/*.
Risk assessment: The commands are limited to local project metadata extraction (git history, file listings) and interaction with a well-known service (GitHub CLI). No user-supplied arguments are interpolated into these shell commands, significantly reducing the risk of command injection.

game-postmortem