The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill is instructed to read the api/src/environments/env.local.yml file to extract the DATABASE_URL. This file typically contains sensitive connection strings, passwords, or access tokens.
[COMMAND_EXECUTION]: Executes shell commands whoami and git config user.name to retrieve the current user's identity for the purpose of generating standardized file paths for the output specifications.
[DATA_EXPOSURE]: The skill ingests content from a wide range of files (Vue components, TypeScript stores, mock data, and DDL schemas) across user-specified paths. This capability allows the agent to access and process sensitive project source code.
[PROMPT_INJECTION]: The workflow involves reading and analyzing untrusted external data (UI prototypes and feature documentation) without explicit boundary markers or instructions to ignore embedded commands. This creates a surface for indirect prompt injection where malicious instructions hidden in the source material could influence the agent's behavior.
Ingestion points: reads from _task-meta.ts, .vue, .mock.ts, .store.ts, and .ts files within a user-provided prototype path, as well as feature documentation folders.
Boundary markers: Absent. The skill does not implement delimiters or warnings to ignore instructions within the read content.
Capability inventory: Performs file reads, executes local shell commands (whoami, git), and writes markdown files to the local filesystem.
Sanitization: No sanitization or validation of the input file content is mentioned before processing or inclusion in the final specification.

peach-gen-spec