The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructions direct the agent to download and execute a remote shell script using the pattern curl -fsSL ... | sh. This occurs in the 'Optional dependency' and 'Step 0 — Bootstrap' sections to install the Higgsfield CLI.
Evidence: curl -fsSL https://raw.githubusercontent.com/higgsfield-ai/cli/main/install.sh | sh in SKILL.md.
[EXTERNAL_DOWNLOADS]: The skill requires downloading a binary (@higgsfield/cli) from a remote repository. While this is the intended purpose of the skill, executing unverified code from the internet poses a security risk.
[COMMAND_EXECUTION]: The skill extensively uses the Bash tool to execute higgsfield CLI commands, perform account status checks, and manage media uploads. It also uses shell pipes (echo "..." | higgsfield ...) which can be exploited if user input is not properly sanitized.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests data from untrusted external URLs to fetch product information.
Ingestion points: The command higgsfield marketing-studio products fetch --url <url> in SKILL.md and references/marketing-products.md reads content from arbitrary web pages.
Boundary markers: Absent. The instructions do not specify delimiters or warnings to ignore instructions embedded within the fetched URL content.
Capability inventory: The skill has access to Bash, Read, Write, and Edit tools, allowing it to execute subprocesses and modify local files based on instructions received.
Sanitization: Absent. There is no mention of validating or escaping the content retrieved from the external URLs before it is processed by the agent or the CLI.

higgsfield-generate