The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/lib/utils.py executes the git blame command using subprocess.run to determine the author of specific history files. The command is invoked with a list of arguments, which is a safe practice to prevent shell injection vulnerabilities.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests and summarizes data from .specstory/history/*.md files, which contain logs of previous AI interactions. These summaries are then presented to the agent, which is instructed to generate a personalized commentary. This creates a surface for indirect prompt injection where malicious instructions in the history files could influence the agent's behavior.\n
Ingestion points: Markdown files in the .specstory/history/ directory are read and parsed in scripts/lib/parser.py.\n
Boundary markers: No security delimiters or instructions to ignore embedded commands are used when presenting the summarized content.\n
Capability inventory: The skill can execute local commands (git) and write files to the disk.\n
Sanitization: The skill performs basic removal of HTML-like tool tags in scripts/lib/report.py but does not sanitize the natural language content for potential prompt injection attacks.

specstory-yak