The Agent Skills Directory

[COMMAND_EXECUTION]: The skill workflow involves executing several local Python scripts, including scripts/update_state.py, scripts/generate_v4_clock.py, scripts/render_snapshot_report.py, and scripts/animate_transition.py, to process ingested data and generate output assets.\n- [DATA_EXFILTRATION]: Reports are automatically transmitted to external services, including a specific hardcoded Discord channel (1444189585373663417) and configured Bluesky accounts. These destinations are prioritized for autonomous delivery as defined in the skill logic.\n- [PROMPT_INJECTION]: \n
The instructions mandate the use of a specific persona ("Good Morning, Cyber Soldiers") characterized by theatrical and urgent language, which constrains the agent's natural output profile.\n
The skill instructs the agent to bypass typical user confirmation for social media posting by directing it to publish to all configured targets by default unless explicitly restricted.\n- [EXTERNAL_DOWNLOADS]: The skill uses websearch and webfetch to retrieve information from arbitrary internet sources, including news articles and PDFs, which are then integrated into the internal state model.\n- [PROMPT_INJECTION]: There is a clear attack surface for indirect prompt injection because the skill ingests untrusted maritime and energy market signals from the web and uses them to generate reports that are then automatically published.\n
Ingestion points: Signal extraction phase in SKILL.md using websearch and webfetch tools.\n
Boundary markers: None identified; untrusted web content is not isolated from the agent's instructional context.\n
Capability inventory: The agent has the authority to execute local scripts and perform writes to external social media via discord and bluesky tools.\n
Sanitization: No validation or sanitization of the external web signals is mentioned before the data is processed or shared publicly.

hormuz-risk-clock