The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands using bash to create symbolic links (ln -s) and manipulate text in workflow files (grep, tail). It also performs file write operations to append new job definitions to .github/workflows/publish.yml. These jobs are configured to access the PUBLISH_TOKEN secret, which is a sensitive CI/CD operation.
[EXTERNAL_DOWNLOADS]: Uses the bunx runner to execute the @plaited/development-skills package for validating skill structures during the scaffolding process.
[PROMPT_INJECTION]: Identifies a potential surface for indirect prompt injection through the use of untrusted user input (the skill name).
Ingestion points: User input for the <skill-name> placeholder is used in various file paths, shell commands, and YAML configurations in SKILL.md and publish.yml.
Boundary markers: None identified. The skill name is directly interpolated into commands and configuration files without delimiters.
Capability inventory: Utilizes Bash for command execution and Write/Edit for modifying sensitive project files like GitHub workflows.
Sanitization: None identified. The instructions do not specify any validation or sanitization for the skill name string before it is used in shell scripts or YAML blocks.

add-skill