The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a significant surface for indirect prompt injection (Category 8) by interpolating untrusted external data into agent prompts.
Ingestion points: The skill ingests untrusted data from bug reports (bug-investigation.yml), feature requests (feature-build.yml), and web research sources including Reddit, Hacker News, and Twitter (research-deep.yml).
Boundary markers: Absent. Untrusted content is interpolated into templates using simple variables like {{bug_description}} or {{web_findings}} without structural delimiters or specific instructions to treat the content as data rather than instructions.
Capability inventory: The orchestration system has extensive capabilities including file system writes, and execution of shell commands via git, npm, and npx across all workflow files.
Sanitization: Absent. The skill relies on subsequent LLM "Reviewer" or "Verifier" steps to catch errors, which are themselves vulnerable to the same injection techniques.
[COMMAND_EXECUTION]: The skill performs extensive shell command execution to manage the development lifecycle.
Evidence: Commands such as git checkout, git commit, git add, npm ci, npx tsc, and npx vitest are used in workflows like 'overnight-batch.yml' and 'feature-build.yml' to automate testing and version control.
[EXTERNAL_DOWNLOADS]: The skill downloads external software dependencies as part of its migration and environment setup tasks.
Evidence: Use of npm install and npm ci in 'migration.yml' and 'overnight-batch.yml' to fetch packages from the official npm registry, which is a well-known service.

task-commander