The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection because it ingests and processes untrusted project data (such as descriptions, risk mitigation plans, and issue resolutions) from ServiceNow tables.
Ingestion points: Data is retrieved from pm_project, pm_project_task, risk, issue, and sys_journal_field tables as described in SKILL.md (Steps 1-4).
Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are utilized when interpolating retrieved data into the report templates in Steps 5-8.
Capability inventory: The skill possesses the ability to write to the ServiceNow environment via SN-Create-Record and has access to a Bash shell.
Sanitization: There is no evidence of filtering, validation, or escaping of the external content before it is processed by the agent.
[DATA_EXFILTRATION]: The skill retrieves sensitive project metrics, including financial data (budget_cost, actual_cost), resource allocation, and internal risk assessments. While this is necessary for its primary function, it involves handling sensitive organizational information that could be targeted for exfiltration if the agent's behavior is compromised.
[COMMAND_EXECUTION]: The skill explicitly includes Bash as an allowed native tool in the frontmatter. While no malicious scripts are present in the provided documentation, the availability of a shell environment increases the potential impact if the agent were manipulated through indirect injection points.

cwm-doc-generation