The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a bundled bash script (scripts/crawlora.sh) to interact with the vendor API. The script implements multiple security controls, including a fixed base URL, a strict whitelist of allowed API paths, and validation of HTTP methods. It further sanitizes path inputs to prevent traversal and blocks the use of characters like '@' in query arguments, mitigating the risk of local file disclosure through curl.
[CREDENTIALS_UNSAFE]: The skill handles the required CRAWLORA_API_KEY following security best practices. The script validates the key's format and uses a temporary curl configuration file with restricted filesystem permissions (chmod 600) to provide the API key as a header, ensuring the secret is not exposed in the system's process listings or shell history.
[INDIRECT_PROMPT_INJECTION]: The skill's functionality involves scraping and analyzing data from external websites, which introduces a potential surface for indirect prompt injection attacks.
Ingestion points: External website content is retrieved through the /web/scrape and /web/techstack endpoints as described in SKILL.md.
Boundary markers: The prompt instructions do not specify the use of delimiters to isolate the untrusted web data from the agent's core instructions.
Capability inventory: The skill maintains network access via the scripts/crawlora.sh utility calling the Crawlora API.
Sanitization: There is no explicit sanitization step described for the markdown content retrieved from external sources before it is processed by the agent.

techstack-prospecting