The Agent Skills Directory

[SAFE]: The skill is a diagnostic and educational utility designed to improve agent reliability by identifying context management issues.
[INDIRECT_PROMPT_INJECTION]: The skill's ingestion surface was evaluated for risks involving untrusted data processing.
Ingestion points: The analyze_agent_context function and PoisoningDetector class in scripts/degradation_detector.py ingest arbitrary context strings for analysis.
Boundary markers: No specific delimiters or safety instructions are used within the analysis script.
Capability inventory: No dangerous capabilities such as subprocess calls, eval/exec, file-writing, or network operations are present in any script.
Sanitization: The input is processed exclusively through string manipulation and regular expression matching for health scoring, presenting no path for execution or exploitation.
[COMMAND_EXECUTION]: Scripts and instructions were analyzed for system command execution. No instances of shell commands, subprocess spawning, or OS-level manipulation were detected.
[EXTERNAL_DOWNLOADS]: Links provided in the documentation refer to public GitHub repositories and research papers for reference purposes. No automated downloads or script execution from remote sources occur.

context-degradation