The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it fetches and processes content from arbitrary external websites.
Ingestion points: Data is retrieved from user-provided URLs via curl in scripts/basic-extract.sh and through the Firecrawl API in scripts/firecrawl-extract.sh.
Boundary markers: The skill lacks boundary markers or instructions to ignore embedded commands; instructions in SKILL.md specifically require the agent to "Preserve exact language" and "Quote, do not paraphrase."
Capability inventory: The agent can execute shell scripts, perform network operations, and write files to the local workspace.
Sanitization: While HTML tags are stripped, the text content is not sanitized or filtered for instructions that might target the LLM.
[EXTERNAL_DOWNLOADS]: The skill performs network requests to the Firecrawl service (api.firecrawl.dev) and various target websites to extract brand DNA. It also downloads binary assets such as logos and screenshots from these external URLs.
[COMMAND_EXECUTION]: The skill executes local shell scripts (scripts/basic-extract.sh and scripts/firecrawl-extract.sh) that utilize standard system tools like curl, jq, grep, sed, and python3 for data processing.
[DATA_EXFILTRATION]: The skill utilizes a FIRECRAWL_API_KEY for API requests to api.firecrawl.dev. It attempts to load this credential from environment variables and common local configuration paths such as .env and ~/.openclaw/credentials/firecrawl.env.

site-extract