The Agent Skills Directory

[SAFE]: The skill operates by inspecting repository manifests (e.g., package.json, pyproject.toml) and directory structures to generate grounded instructions. This evidence-based approach prevents the invention of arbitrary or malicious commands.
[SAFE]: Robust security guidelines are integrated into the core instructions. The agent is explicitly instructed to ignore any prompts or commands embedded in the files it processes, effectively mitigating indirect prompt injection attempts.
[SAFE]: The use of the execute capability is strictly scoped to local verification of existing project scripts. The instructions include mandatory gates that block destructive actions, credential usage, network exfiltration, and production deployments.
[SAFE]: The validation and testing workflow utilizes local scripts (skills/skill-tester/scripts/validate-skills-corpus.mjs) to ensure artifact integrity without relying on untrusted external resources.
[SAFE]: No evidence of obfuscation, hidden URLs, persistence mechanisms, or unauthorized privilege escalation was found in the skill's instructions or supporting rules.

agent-md-maker