The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external websites (such as robots.txt content, HTTP headers, and body previews) which could contain malicious instructions designed to subvert agent behavior.
Ingestion points: Data is fetched by scripts/scan_site.sh and processed by scripts/triage_scan.py and scripts/render_plan.py (specifically via scan.json and fixes.md).
Boundary markers: The skill includes an explicit "Prompt Injection Boundary" section in SKILL.md which mandates that scanner content be treated as data/content rather than instructions, and provides specific rules for handling phrases like "ignore previous instructions".
Capability inventory: The skill can write to the local file system (agent-ready-plan.md), perform network operations via curl, execute local shell and Python scripts, and delegate issue creation to another skill (plan-to-issues).
Sanitization: The scripts/render_plan.py script uses a clean() function to sanitize scanner-derived strings by collapsing newlines, stripping markdown heading markers (#), and escaping table delimiters (|) to prevent site content from forging plan structure.
[EXTERNAL_DOWNLOADS]: The skill makes network requests to isitagentready.com to perform scans and retrieve remediation prose. This is the primary function of the skill and is protected by a mandatory human approval gate (G1) before any data is sent.
[COMMAND_EXECUTION]: The skill executes several internal scripts (scan_site.sh, triage_scan.py, render_plan.py) and utilizes standard CLI tools including bash, python3, git, and the GitHub CLI (gh) for repository management and dependency verification.

website-agent-readiness