The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from the Xiaohongshu platform, creating a vulnerability surface for indirect prompt injection.
Ingestion points: tools/xhs-content-ops/scripts/fetch_note_content.ts downloads full note text, descriptions, and comments. tools/xhs-engagement/scripts/xhs_engagement.py scrapes note titles from the creator backend.
Boundary markers: The workflows and scripts do not implement explicit delimiters or "ignore embedded instructions" warnings when processing the ingested note content.
Capability inventory: The agent has the capability to write to the local file system (xhs/dna/, xhs/outputs/), execute shell commands (camoufox-cli, curl), and perform network operations to XHS APIs (publish_xhs.py).
Sanitization: The skill performs minimal sanitization, such as whitespace normalization in xhs_engagement.py and regex-based topic extraction in publish_xhs.py.
[COMMAND_EXECUTION]: The skill frequently executes external commands and local scripts to perform its tasks.
Evidence: tools/xhs-engagement/scripts/xhs_engagement.py uses subprocess.run to call camoufox-cli for browser automation and update-metrics.sh for database updates.
Evidence: tools/xhs-content-ops/scripts/fetch_note_content.ts executes curl via child_process.execFile to resolve short links and download images.
Evidence: tools/xhs-publish/scripts/login-and-verify.ts executes camoufox-cli to handle SSO redirection and session exports.

expert-xhs