The Agent Skills Directory

[SAFE]: The skill is a legitimate technical resource for context engineering. It focuses on efficiency, cost reduction, and performance optimization for LLM-based agents.
[COMMAND_EXECUTION]: The skill includes Python scripts (scripts/compaction.py) that implement token estimation, regex-based summarization, and state management for tool observations. These operations are performed locally using standard Python libraries (re, hashlib, time) and do not involve shell command execution or dangerous subprocesses.
[EXTERNAL_DOWNLOADS]: No external network requests or remote script downloads were detected. The Python code is self-contained and uses internal logic for its demonstrations.
[DATA_EXFILTRATION]: There is no evidence of sensitive file access or network transmission. The ObservationStore class implements a local hashing mechanism for reference IDs but does not send data externally.
[PROMPT_INJECTION]: The instructions in SKILL.md are purely technical and guide the agent on how to manage its own context window. There are no attempts to bypass safety filters or override system instructions for malicious purposes.
[OBFUSCATION]: The content is presented in clear markdown and standard Python code. No hidden characters, Base64-encoded strings, or homoglyphs were found.

context-optimization