The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides a local shell script, scripts/render-dashboard.sh, which the agent is instructed to execute to generate a visual dashboard. The script includes a structural validation layer using jq to ensure the integrity of the audit data before rendering. This is a functional component of the skill's reported purpose.\n- [EXTERNAL_DOWNLOADS]: The dashboard template (assets/dashboard-template.html) references the Chart.js library via JSDelivr (cdn.jsdelivr.net). This is a recognized well-known service for hosting front-end assets and is considered safe per the trust scope rules.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from audited websites and local files. It implements an evidence chain to mitigate risks: 1) Ingestion points are the WebFetch tool and local file reads. 2) Boundary markers and explicit instructions in rules/validation.md warn the agent to treat external content as text. 3) Capability inventory includes local file writes and shell execution. 4) Sanitization is provided via the escapeHtml function in the dashboard template and a custom JSON serializer (jser) in the shell script's validation logic.\n- [DATA_EXPOSURE]: The skill reads project metadata and technical configurations to perform audits, saving results to the .hyper/seo/ directory. These actions are transparently documented as the primary intent of the tool and do not involve unauthorized harvesting or exfiltration of sensitive credentials.

hyper-seo