The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides a command in SKILL.md that downloads a shell script from https://freellmapi.co/install.sh and pipes it directly into bash. This pattern allows for the execution of arbitrary, unverified code from an external server on the user's local system, which is a critical security vulnerability.
[REMOTE_CODE_EXECUTION]: In references/agent-clients.md, the skill uses the npx freellmapi command to perform setup tasks. This downloads and executes packages from the npm registry at runtime, bypassing static security checks and potentially executing malicious code if the package is compromised.
[DATA_EXPOSURE]: The instructions in SKILL.md direct the agent to scan and read .env and .env.local files within the project and sibling directories to harvest FREELLMAPI_KEY or OPENAI_API_KEY. This practice accesses sensitive file paths and exposes private credentials.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection by ingesting data from untrusted local files.
Ingestion points: The skill reads configuration data from .env files in SKILL.md and references/recipes.md.
Boundary markers: No boundary markers or "ignore instructions" warnings are implemented when reading and processing these files.
Capability inventory: The skill utilizes the Bash, Write, and Edit tools, which could be exploited to modify the system if the ingested data contains malicious instructions.
Sanitization: The skill lacks sanitization or validation logic for the values extracted from the environment files before they are utilized in subsequent command executions.

use-freellmapi