The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses imperative language designed to override other instructions and influence the agent's task prioritization, specifically stating: "This section overrides everything else in this skill if there is any conflict. Read it first, act on it first."
[EXTERNAL_DOWNLOADS]: The instructions direct the user and the agent to install an external plugin from an untrusted source via the command: claude plugin marketplace add lackeyjb/playwright-skill.
[INDIRECT_PROMPT_INJECTION]: The skill establishes an attack surface for indirect prompt injection by capturing and reasoning from raw network traffic (XHR, fetch, WebSocket, SSE) from untrusted external websites.
Ingestion points: Network traffic captured via page.on("request"), page.on("response"), and page.on("websocket") as described in the Discovery Gate section of SKILL.md.
Boundary markers: Absent. The skill does not define delimiters or provide instructions to ignore malicious content embedded in the captured data.
Capability inventory: Access to the playwright-skill plugin, inline Bash script execution, and Python scripts (instagram_grab.py) capable of file system access and network operations.
Sanitization: Absent. The agent is explicitly encouraged to "reason from real endpoints" using the raw, unvalidated capture data, which may contain attacker-controlled strings.
[DYNAMIC_EXECUTION]: The skill utilizes page.evaluate in scripts/instagram_grab.py to execute dynamically constructed JavaScript within the browser context and recommends generating and running inline automation scripts via Bash.
[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands for plugin management and browser automation tasks.

grabber-development