The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use the codex CLI with the --yolo flag, explicitly described as having "no sandbox, no approvals — fastest." This promotes the bypass of local system safety mechanisms.
[CREDENTIALS_UNSAFE]: The skill provides the exact path and command (op read "op://DeLoSecrets/Jules/API Key") to programmatically retrieve sensitive API keys from a 1Password vault. While using a secret manager is a security best practice for humans, providing direct access paths to an AI agent for credential harvesting increases the risk of unauthorized exposure.
[DATA_EXFILTRATION]: The orchestration patterns involve sending local repository content to multiple external AI providers (OpenAI, Anthropic, Google, Kimi, Augment) for processing. While necessary for the skill's purpose, this represents a significant data export surface.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection by design, as it facilitates the processing of untrusted repository code by multiple sub-agents equipped with shell execution capabilities.
Ingestion points: Local repository files accessed via workdir paths in various exec and sessions_spawn commands.
Boundary markers: None (the skill does not provide instructions to wrap or ignore instructions found in the processed code).
Capability inventory: Full shell command execution (exec) with PTY support and the ability to spawn new agent sessions (sessions_spawn).
Sanitization: No sanitization, escaping, or validation of the code content is performed before it is processed by the LLM agents.

coding-strategy