The Agent Skills Directory

[COMMAND_EXECUTION]: The /seo flow sync command is configured to execute a local Python script sync_flow.py using the claude-seo run command. As this script is not included in the provided skill files, its logic is unverifiable and could potentially execute arbitrary code if a malicious version of the file is present on the host system.
[EXTERNAL_DOWNLOADS]: The skill's synchronization logic involves downloading and updating prompt files from an external repository (github.com/AgriciDaniel/flow). While this repository is related to the skill's stated framework, downloading content from unverified third-party sources into the skill's operational environment poses a supply chain risk.
[REMOTE_CODE_EXECUTION]: The orchestration of sync_flow.py to fetch and overwrite the skill's own prompt files constitutes a remote code execution surface. Since the agent subsequently loads and obeys these downloaded instructions, an attacker compromising the source repository could influence the agent's behavior globally.
[DATA_EXFILTRATION]: The orchestration logic and error handling reference the use of the GitHub CLI (gh auth login) and the GitHub API. Instructions to authenticate or interact with credential-managing tools can be used as a vector to capture or expose sensitive access tokens if the process is manipulated.
[PROMPT_INJECTION]: The skill processes untrusted external data, specifically user-provided URLs and topics, which are interpolated into SEO analysis prompts. This creates an indirect prompt injection surface where malicious content on a target website could attempt to override the agent's instructions or exfiltrate data from the current session.

seo-flow