The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads and processes script content from remote ServiceNow tables. Malicious instructions embedded within these remote scripts could potentially influence the agent's behavior during synchronization.
Ingestion points: SN-Query-Table and SN-Get-Record tools (in SKILL.md) are used to fetch script content from tables such as sys_script and sys_script_include.
Boundary markers: There are no explicit instructions or delimiters used to tell the agent to ignore instructions contained within the fetched script data.
Capability inventory: The skill utilizes Write for local file access, Bash for shell command execution, and SN-Update-Record for modifying remote ServiceNow data.
Sanitization: No sanitization or validation mechanisms are described to verify the content of the scripts before they are processed or written to disk.
[COMMAND_EXECUTION]: The instructions utilize local shell commands for project initialization and version control management.
Evidence: Commands such as mkdir, git init, git add, git commit, git remote add, and chmod are documented in the workflow.
[EXTERNAL_DOWNLOADS]: The skill facilitates the retrieval of source code from external ServiceNow instances via specialized tools and REST APIs.
Evidence: The skill uses SN-Sync-Script-To-Local, SN-Get-Record, and REST endpoints like /api/now/table/sys_script to pull data from the instance.

script-sync