The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the sau CLI tool to perform actions such as logging in, uploading videos, and deleting posts.
Evidence: The script scripts/examples/instagram_cli_template.py uses subprocess.run(command, check=True) to execute these operations. The implementation uses list-based arguments, which is a secure practice that avoids spawning a shell and prevents most command injection vectors.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process user-provided metadata, such as titles, descriptions, and tags, which are then passed as arguments to a command-line interface.
Ingestion points: User-supplied strings for --title, --desc, --tags, and --note as described in SKILL.md and the example scripts.
Boundary markers: The provided scripts do not include explicit sanitization or delimiters for these strings, relying on the agent's interpolation logic.
Capability inventory: The skill has the capability to execute shell commands and interact with the Instagram platform via the sau CLI.
Sanitization: No explicit sanitization is shown in the examples, although the Python subprocess list usage provides structural protection against argument splitting.
[CREDENTIALS_UNSAFE]: The skill manages authentication state by storing session cookies in local JSON files.
Evidence: SKILL.md identifies that account files are saved as cookies/instagram_creator.json. While the skill correctly instructs the user not to commit sensitive account status or screenshots to version control, the existence of these cookie files represents a sensitive data surface on the local filesystem.

instagram-upload