The Agent Skills Directory

REMOTE_CODE_EXECUTION (CRITICAL): The installation instructions command the user to pipe a remote script from an untrusted domain directly into bash (curl -sSL https://canifi.com/skills/bitbucket/install.sh | bash). This allows the remote server to execute arbitrary commands on the host machine without verification.
EXTERNAL_DOWNLOADS (HIGH): The skill references multiple external scripts from canifi.com, which is not a verified or trusted source. This includes the main skill installer and the canifi-env configuration utility.
COMMAND_EXECUTION (HIGH): The skill relies on the execution of local shell commands for environment configuration and tool operations, providing a pathway for privilege escalation if the downloaded scripts are malicious.
PROMPT_INJECTION (LOW): The skill possesses a significant attack surface for indirect prompt injection. It ingests data from Bitbucket (Pull Requests, issues, commit messages, and pipelines). An attacker could place malicious instructions in a repository's metadata or comments that the AI agent might follow when summarizing activity.
Ingestion points: bitbucket.org via Playwright MCP (PRs, Issues, Pipelines, Activity feeds).
Boundary markers: Absent. The skill does not define specific delimiters for external content.
Capability inventory: Shell command execution (via canifi-env), browser automation via Playwright, and file system access to the .canifi directory.
Sanitization: Absent. No evidence of content filtering or safety instructions regarding external repository data.

bitbucket