The Agent Skills Directory

[PROMPT_INJECTION]: No malicious patterns intended to bypass safety guardrails or override core agent behavior were detected. The instructions use standard, benign language to define the skill's logic.
[DATA_EXPOSURE]: There is no evidence of hardcoded credentials, sensitive file path access (e.g., .ssh, .aws), or unauthorized data collection.
[DATA_EXFILTRATION]: The skill does not perform network operations to external domains for the purpose of exfiltration. All referenced URLs (github.com, spaceappschallenge.org) are well-known and relevant to the skill's purpose.
[REMOTE_CODE_EXECUTION]: No patterns for remote script execution or unauthorized command execution were found. References to other skills by the same author ('samber') represent modular skill design rather than RCE.
[OBFUSCATION]: The skill body and metadata contain no encoded content, zero-width characters, or homoglyph substitutions designed to hide malicious intent.
[DYNAMIC_CONTEXT_INJECTION]: The skill does not utilize dynamic context injection or shell-command placeholders in its configuration.
[INDIRECT_PROMPT_INJECTION]: While the skill ingests external data (user interview answers and previous rules), it lacks the dangerous capabilities (such as file-writing or network exfiltration) necessary to exploit this attack surface.

hackathon-brief-design