The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the op command-line tool and tmux for session orchestration and secret retrieval.
The workflow in SKILL.md uses op signin, op whoami, and op vault list within a managed tmux session to handle authentication challenges.
[DATA_EXFILTRATION]: The skill is designed to access and handle sensitive credentials from 1Password vaults.
Examples in references/cli-examples.md demonstrate reading passwords and private keys (op read) and writing secrets to disk (op read --out-file).
These behaviors are the primary intended purpose of the skill and are mitigated by guardrails prohibiting the pasting of secrets into logs or chat.
[EXTERNAL_DOWNLOADS]: The skill provides metadata and instructions for installing the official 1Password CLI.
The SKILL.md frontmatter specifies the 1password-cli Homebrew formula for installation.
[INDIRECT_PROMPT_INJECTION]: The skill processes vault paths and secret identifiers which may be sourced from untrusted data.
Ingestion points: Secret paths (e.g., op://app-prod/db/password) and environment variable names used as arguments for the op tool.
Boundary markers: None present in the instructions to separate user-provided paths from system commands.
Capability inventory: Capability to write secrets to files (op inject, op read --out-file) and execute subprocesses with injected secrets (op run) as shown in references/cli-examples.md.
Sanitization: No explicit sanitization or validation of the secret paths or template inputs is mentioned.

1password