The Agent Skills Directory

[COMMAND_EXECUTION]: The SKILL.md file instructs the agent to execute a Python script (template_scaffolder.py) located in a parent directory (../atlassian-templates/scripts/). Referencing and executing code outside of the skill's own directory structure introduces an unverifiable dependency that could lead to unexpected code execution if the external environment is compromised or improperly configured.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection by processing external data and using it to perform operations in Confluence.
Ingestion points: The scripts scripts/space_structure_generator.py and scripts/content_audit_analyzer.py ingest user-supplied data from team_info.json and pages.json respectively.
Boundary markers: The skill does not employ delimiters or explicit instructions to ignore potentially malicious commands embedded within the processed data.
Capability inventory: The skill is capable of modifying Confluence content through the mcp__atlassian__createConfluencePage and mcp__atlassian__updateConfluencePage tools.
Sanitization: The Python scripts extract fields such as titles and descriptions from the JSON input and process them without visible sanitization or validation to prevent prompt-based manipulation of the agent's logic.

confluence-expert