The Agent Skills Directory

[Prompt Injection] (SAFE): The skill focuses on technical documentation for OpenCode agents. No patterns of instruction overrides, role-play injections, or system prompt extraction were found.
[Data Exposure & Exfiltration] (SAFE): No hardcoded credentials or network exfiltration patterns are present. The package_skill.sh script demonstrates a high level of security awareness by maintaining a blacklist of sensitive file patterns (e.g., .ssh, .env, .pem, credentials*) to prevent accidental data exposure during the packaging process.
[Remote Code Execution] (SAFE): The skill does not perform any network operations or download external scripts. All operations are local.
[Command Execution] (SAFE): Shell scripts (init_skill.sh, validate_skill.sh, package_skill.sh) utilize standard Unix utilities (awk, sed, zip, find). User-provided inputs, such as skill names, are strictly validated using the regex ^[a-z0-9]+(-[a-z0-9]+)*$ before being processed, effectively preventing command injection.
[Privilege Escalation] (SAFE): No use of sudo or modification of system-level permissions was found. The use of chmod +x is restricted to locally scaffolded scripts within the skill directory.
[Persistence Mechanisms] (SAFE): The skill does not attempt to modify shell profiles, cron jobs, or any other persistence vectors.
[Obfuscation] (SAFE): All documentation and scripts are provided in clear, human-readable text without any encoding or hidden characters.

build-skill