The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill analyzes external data from user codebases (e.g., package.json, requirements.txt, go.mod) to determine deployment configurations, which represents a vulnerability surface.
Ingestion points: The codebase analysis process described in references/codebase-analysis.md targets multiple framework-specific dependency and configuration files.
Boundary markers: Absent; the instructions do not specify any delimiters or safety warnings to ignore instructions embedded in the analyzed project files.
Capability inventory: The skill can perform file system operations (writing render.yaml), execute shell commands (git, render, curl), and invoke MCP tools for cloud resource creation and configuration.
Sanitization: Absent; there is no explicit instruction to sanitize or validate the content of the analyzed files before using them to generate deployment manifests.
[PRIVILEGE_ESCALATION]: The skill instructions advise the agent to request require_escalated sandbox permissions if network operations are blocked by the environment. While intended for legitimate cloud deployment tasks, this is an explicit request for elevated platform privileges.
[REMOTE_CODE_EXECUTION]: The skill contains a pattern for downloading and executing a remote shell script (curl | sh) for the installation of the Render CLI. This script originates from the official render-oss GitHub organization, which is a well-known service.
[EXTERNAL_DOWNLOADS]: The skill facilitates the download of the Render CLI and references multiple external resources from the Render platform (mcp.render.com, dashboard.render.com).
[METADATA_POISONING]: The skill's author metadata in the YAML frontmatter (github.com/openai/skills) is inconsistent with the provided author context (tech-leads-club).
[COMMAND_EXECUTION]: The skill involves executing local commands, including git operations for repository management and Render CLI commands for blueprint validation and service discovery.

render-deploy