The Agent Skills Directory

[Prompt Injection] (SAFE): The skill uses natural language instructions to guide the agent. No malicious override patterns, role-play jailbreaks, or system prompt extraction attempts were detected. Instructions like 'IMPORTANT' are used in a benign, instructional context.
[Data Exposure & Exfiltration] (SAFE): No hardcoded credentials, sensitive file paths, or network operations (curl, wget, fetch) were found. The URLs in skillpack.json are for metadata and documentation purposes only.
[Obfuscation] (SAFE): No Base64 encoding, zero-width characters, homoglyphs, or hex escapes were detected in the text or metadata.
[Unverifiable Dependencies & Remote Code Execution] (SAFE): The skill does not include any package manifest files (package.json, requirements.txt) or commands to install external software. No remote script execution patterns were identified.
[Privilege Escalation] (SAFE): No commands involving sudo, chmod, or system configuration modifications are present.
[Persistence Mechanisms] (SAFE): No attempts to modify shell profiles, cron jobs, or startup services were found.
[Metadata Poisoning] (SAFE): Metadata in skillpack.json and the YAML frontmatter in SKILL.md is accurate and describes the skill's purpose without deceptive instructions.
[Indirect Prompt Injection] (SAFE): While the skill ingests user-provided context about products and customers, it possesses no exploitable capabilities (such as file writing, shell execution, or network calls) that could be triggered by malicious data within that context.
[Time-Delayed / Conditional Attacks] (SAFE): No logic exists that gates behavior based on time, date, or specific environment triggers.
[Dynamic Execution] (SAFE): The skill does not generate, compile, or execute code at runtime. It produces structured text (Markdown) as its primary output.

writing-north-star-metrics