The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-edited text and interpolates the content directly into analysis prompts without boundary markers or sanitization, creating a surface for embedded instructions to influence the agent.
Ingestion points: The pattern-analysis module in modules/pattern-analysis.md reads post_review_text and post_edit_text from user files.
Boundary markers: The analysis prompt lacks delimiters or instructions to ignore embedded commands within the text being compared.
Capability inventory: The skill has the ability to execute shell commands (bash) and write to the local file system (cp) via the snapshot-management module and Step 5 of the main skill.
Sanitization: There is no evidence of sanitization or validation of the text content before it is processed by the agent.
[COMMAND_EXECUTION]: The skill utilizes shell commands to manage its internal state and snapshot history.
Evidence: SKILL.md and modules/snapshot-management.md include bash snippets for directory creation, file copying (cp), and manifest updates, using variables like {name} for file paths.

voice-learn