The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill explicitly references and checks for the existence of Telegram session files located at ~/.local/share/gramjs/<profile>.session. These files contain sensitive authentication tokens and session data required to access and control a Telegram account.
[INDIRECT_PROMPT_INJECTION]: The skill contains a 'Self-Evolving Skill' section and 'Post-Execution Reflection' instructions that direct the agent to 'fix this file immediately' if errors occur or workarounds are needed. This creates a mechanism where external output from the Telegram CLI or unexpected runtime behavior could influence the agent to persistently modify its own instruction set (SKILL.md), potentially leading to persistent prompt injection or logic manipulation.
[INDIRECT_PROMPT_INJECTION]: The skill accepts user-provided input (usernames, phone numbers, IDs) and interpolates them directly into a Bash heredoc block. There are no explicit instructions for the agent to sanitize or validate these inputs before they are passed to the bun command, creating a potential vector for command injection if the agent does not apply its own safety filters.
[COMMAND_EXECUTION]: The skill executes a local TypeScript script (tg-cli.ts) using the bun runtime via a Bash heredoc. This is a standard execution pattern for this plugin's functionality but involves direct interaction with the host shell.

find-user