The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill instructs users to export a sensitive EVM private key (FITTINGS_PRIVATE_KEY) as a plaintext environment variable to facilitate payments. This practice exposes the private key to any process running in the same environment, presenting a significant risk of fund theft.
[REMOTE_CODE_EXECUTION]: The skill uses curl -sO https://fittings.sh/skill/call.mjs followed by node call.mjs to execute tasks. This pattern downloads an unversioned script from a remote server and executes it locally, allowing the service provider to change the code at any time and execute arbitrary commands on the host system.
[EXTERNAL_DOWNLOADS]: The skill fetches configuration and executable code from fittings.sh. While these are vendor-owned resources, the method of retrieval and execution lacks integrity checks (such as hashes or version pinning).
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted free text of up to 200,000 characters to extract metadata. This represents an ingestion surface where malicious content could be embedded.
Ingestion points: The text parameter in the extract-citations call in SKILL.md.
Boundary markers: None provided in the skill instructions to separate data from instructions.
Capability inventory: The skill uses curl for network access and node for local execution.
Sanitization: No sanitization or validation of the input text or extracted output is described.
[COMMAND_EXECUTION]: The skill relies on shell commands (curl, node, export) to interact with the service and manage sensitive credentials.

fittings-text-extract-citations