The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses local shell commands and scripts to manage coding history and automate its workflows.\n
Evidence: Executes git log to identify session authors and attributes work correctly within a local environment.\n
Evidence: Runs internal Node.js scripts for corpus management and to validate forge plans before they are presented to the user.\n- [PERSISTENCE]: The skill manages its own installation and generates persistent artifacts derived from user history.\n
Evidence: Installs by creating symlinks in standard AI agent configuration directories such as ~/.claude/skills and ~/.agents/skills.\n
Evidence: Automatically generates and stores new SKILL.md files that represent reproducible coding patterns discovered in the user's transcripts.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted transcript data but incorporates robust mitigations to prevent malicious prompt execution.\n
Evidence: Ingests markdown transcripts from history directories which could theoretically contain hidden instructions.\n
Evidence: Features a redactSecrets system using regex patterns to strip AWS keys, GitHub tokens, and other sensitive credentials before data is analyzed by the LLM.\n
Evidence: Workflow scripts for subagents explicitly include security directives to treat transcript text as inert data, preventing the model from obeying instructions found within the mined history.\n- [EXTERNAL_DOWNLOADS]: The skill follows standard installation procedures for the agent ecosystem.\n
Evidence: Recommends installation via npx from the official specstoryai repository, which is the verified vendor for this skill.

lore