The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill instructions explicitly direct the agent to search for and read configuration files which are known to contain sensitive information.
Evidence: In references/analysis-checklist.md, the "Config" section suggests searching for **/.env* files.
Evidence: In references/analysis-process.md, Step 6 (Configuration) instructs the agent to use Glob: **/.env* to locate environment files.
Risk: Reading these files exposes API keys, database credentials, and other secrets to the agent's context, which could lead to accidental exposure or misuse.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from external codebases while possessing high-privilege capabilities, creating an attack surface for malicious code to influence agent behavior.
Ingestion points: The skill utilizes Read, Grep, and Glob tools to examine arbitrary files within a target codebase, as outlined in the Core Workflow of SKILL.md.
Boundary markers: The instructions lack explicit boundary markers or warnings to the agent to treat content from analyzed files as untrusted data or to ignore embedded natural language instructions.
Capability inventory: The skill is granted access to the Read, Grep, Glob, and Bash tools in its configuration (SKILL.md frontmatter).
Sanitization: There are no provided mechanisms for sanitizing, validating, or filtering the content of the files before they are processed by the agent, increasing the risk that malicious comments or documentation could trigger unintended tool usage.

spec-miner