The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The detect subcommand processes files in a user-provided directory to identify build and test commands. This creates a vulnerability surface where content in the scanned repository could manipulate the AI's output to include malicious commands in the generated configuration.
Ingestion points: Source code and manifest files located within the target directory (defaults to cwd) processed by the ouroboros detect command.
Boundary markers: No specific delimiters or instructions to ignore embedded commands are defined for the detection phase.
Capability inventory: The skill generates a .ouroboros/mechanical.toml file containing shell commands (lint, build, test, static, coverage) that are executed during the subsequent evaluation stage.
Sanitization: The skill mentions that commands are "validated," but the specific sanitization logic is handled externally by the ouroboros CLI tool.
[COMMAND_EXECUTION]: The skill uses the Bash environment to execute the ouroboros detect command and the uvx runner.
[EXTERNAL_DOWNLOADS]: The skill uses uvx to download and execute the ouroboros-ai package from the Python Package Index (PyPI). This is a vendor-provided tool associated with the skill's author context.
[DYNAMIC_EXECUTION]: The skill implements a workflow where shell commands are dynamically generated by analyzing a codebase and then stored in a configuration file for later execution.

brownfield