The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-generated annotations from 'tuicr' sessions and treats them as instructions for modifying source code.
Ingestion points: The agent reads review comments using the tuicr review comments command, as described in SKILL.md and references/usage.md.
Boundary markers: Absent. The skill does not define specific delimiters or instructions to prevent the agent from blindly following instructions embedded within the annotation text.
Capability inventory: The skill allows the use of Bash, Read, Edit, Write, Grep, and Glob tools, providing the agent with the ability to perform file system modifications and shell command execution based on the processed annotations.
Sanitization: Absent. The workflow instructs the agent to classify and directly "address" the content of the annotations without validation or escaping.
[COMMAND_EXECUTION]: The scripts/launch-tuicr.sh script dynamically constructs and executes shell commands to launch the TUI in external terminal environments like 'herdr' or 'ghostty'.
Evidence: The script assembles a command string including user-supplied arguments and executes it via terminal workspace tools (herdr pane run or ghostty -e).
Context: This mechanism is utilized to provide the 'tuicr' tool with a functional TTY environment, which is not available in the agent's native execution context. The script employs a shell-quoting helper to mitigate injection risks during command assembly.
[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of the 'tuicr' tool from the official nixpkgs registry.
Evidence: SKILL.md and references/install.md provide instructions for fetching the tool using nix build or nix run commands targeting the nixpkgs#tuicr package.

tuicr