The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The references/cli-tools/doc-search.md file suggests using npx @vedanth/context7 to retrieve library documentation. This command downloads and executes code from the NPM registry provided by a source not recognized as a trusted organization.
[REMOTE_CODE_EXECUTION]: The skill facilitates remote code execution by instructing the agent to use npx for tools such as @vedanth/context7 and skills. While the skills tool is provided by a known trusted entity, the execution of code from unverified sources like @vedanth/context7 introduces a risk of executing malicious logic.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through its data processing workflow.
Ingestion points: It reads product specification files (spec-product.md) and examines the codebase using the cymbal search utility in the planning:10.5 stage.
Boundary markers: There are no explicit delimiters or instructions to prevent the agent from following commands that might be embedded within the project files it reads.
Capability inventory: The skill can perform file operations (writing spec-tech.md), execute shell utilities (jq, awk, grep), and call specialized subagents like the planner.
Sanitization: The skill does not implement sanitization or validation for content extracted from the product specifications before it is used to generate further instructions.
[COMMAND_EXECUTION]: Several bash scripts are used to automate stack detection and tech planning validation. These scripts utilize standard command-line tools (jq, grep, sed, awk) to process content from the project directory, which may include untrusted data.

stelow-workflow-tech-planning