The Agent Skills Directory

[DATA_EXFILTRATION]: The skill manages sensitive configuration and credential files.
Accesses ~/.gemini/config/channels/discord/.env, which is used to store the sensitive DISCORD_BOT_TOKEN credential.
Accesses ~/.gemini/config/channels/discord/access.json, which contains user identifiers (snowflakes) and access control policies.
Note: The skill implements mitigation strategies by masking the token in the status display and using restrictive file permissions.
[COMMAND_EXECUTION]: The skill performs directory and file system permission operations.
Uses mkdir -p to create the required configuration directory structure.
Executes chmod 600 on the .env file to ensure the stored bot token is only accessible by the owner.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied data which represents an attack surface for embedded instructions.
Ingestion points: User input captured via the $ARGUMENTS parameter in SKILL.md when setting the token.
Boundary markers: Absent; the skill treats the input as a literal token string without using delimiters or warnings to ignore embedded content.
Capability inventory: Includes file system writes (write_to_file), reads (view_file), directory creation (mkdir), and permission modification (chmod).
Sanitization: The skill performs whitespace trimming on the input token string.

configure