Security risk report for @forgecat/ayghri_i-have-adhd v0.1.2
Source Integrity
Low
Profile references a legitimate public GitHub repository (ayghri/i-have-adhd) with MIT license and verifiable commit hash.
No typosquatted, obfuscated, or attacker-controlled package dependencies declared; no external binary execution or supply-chain injection vectors.
Agent Intent
Low
Content describes a legitimate output-formatting skill (ADHD-friendly brevity rules) with no instructions to manipulate the AI, exfiltrate data, read credentials, or hide its own instructions.
The skill rules (lead with action, number steps, suppress tangents, etc.) are stylistic guidance for how the agent should format responses—standard productivity advice, not prompt injection or guidance poisoning.
No system prompt leakage, role hijacking, or instructions to fetch external URLs, install malicious dependencies, or weaken security defaults.
Details
Evidence
Rule 1: 'Lead with the next action' — formatting instruction, not manipulation.
Rule 10: 'No preamble, no recap, no closing pleasantries' — style guidance, not malicious intent.
No references to reading ~/.ssh, ~/.aws, .env, or exfiltrating data.
Permissions
Low
No tools or skills declared; the profile is purely instructional markdown defining output-formatting rules.
No file_write, file_delete, shell, web_fetch, or other high/medium-risk tool categories requested.
Authority is limited to shaping the agent's response format, which aligns with the stated purpose.
Details
Evidence
Declared tools section: (none).
Skill description: 'Shape output for a reader with ADHD' — no system-level or file-system access required.
MCP Risk
Low
No MCP servers defined in the profile.
No hidden instructions in tool descriptions, arbitrary binary execution, or unrestricted network/filesystem access.
ForgeCat