The Agent Skills Directory

[PROMPT_INJECTION]: No attempts to override safety filters, bypass constraints, or extract system prompts were detected. The skill uses standard instructional language for its intended purpose.
[DATA_EXFILTRATION]: No network access commands (curl, wget) or sensitive file path references (e.g., .ssh, .env) were found. The skill operates solely on text provided in the user input.
[REMOTE_CODE_EXECUTION]: No patterns of remote script execution or package installation were identified. The skill's frontmatter explicitly disables model invocation, preventing the AI from calling external tools or code execution environments.
[COMMAND_EXECUTION]: No shell commands or system-level scripts are included in the instructions. The skill is purely generative text processing.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill ingests untrusted data via the kud_charts and competency_structure variables in SKILL.md.
Boundary markers: The prompt uses clear structural headers to separate instructions from data, though it does not include explicit delimiters for the input variables.
Capability inventory: The skill has no capability for tool execution, file access, or network calls, as restricted by the disable-model-invocation: true setting.
Sanitization: No explicit sanitization of input data is performed.
Conclusion: While the skill processes untrusted input, the lack of executable capabilities makes the indirect prompt injection surface non-exploitable.

developmental-progression-synthesis