The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's YAML profile structure supports a hooks field that allows the execution of arbitrary shell commands at various stages of the transfer process (on_start, on_complete, on_error, on_retry).
Evidence: hooks: { on_start: "/opt/scripts/notify_start.sh", on_complete: "/opt/scripts/notify_complete.sh {bytes} {duration}", ... } in SKILL.md.
[DYNAMIC_EXECUTION]: The provided Node.js automation wrapper (FastCopySync class) is vulnerable to shell command injection. It constructs a command string by joining arguments and executes it using execSync without sanitizing input parameters.
Evidence: const cmd = args.join(' '); ... const output = execSync(cmd, { encoding: 'utf8' }); in the Node.js example.
[PRIVILEGE_ESCALATION]: The installation and troubleshooting sections encourage the use of sudo for administrative tasks such as moving binaries to /usr/local/bin/ and executing transfers, which escalates the risk of any command execution vulnerability.
Evidence: sudo mv fastcopy /usr/local/bin/ and sudo fastcopy /source /dest in SKILL.md.
[INDIRECT_PROMPT_INJECTION]: The skill establishes an attack surface where untrusted configuration data (YAML profiles) is ingested and used to trigger high-privilege operations and arbitrary shell hooks.
Ingestion points: YAML profile files loaded via the --profile flag.
Boundary markers: None present; the skill does not implement delimiters or warnings for embedded instructions in profiles.
Capability inventory: Arbitrary shell execution via hooks, file system writes, and network operations via curl in examples.
Sanitization: Absent in the provided Node.js wrapper example.
[EXTERNAL_DOWNLOADS]: The documentation references downloading the utility from an unspecified "official source or repository" without providing verifiable URLs or checksums, which could lead users to untrusted binaries.
Evidence: "Download from official source or repository" in the Windows installation section.

fastcopy-file-transfer-utility