tokenchronicle-insights
Summarize private TokenChronicle usage reports and memory candidates while preserving source provenance and user approval boundaries.
它会碰到什么
这一栏是扫描器报的事实,不是结论。命中多不等于有毒(安全工具、规则库、示例脚本本来就会包含危险写法),命中少也不等于干净。它和你手上的凭据、文件、网络有什么关系,需要你自己看。
技能内容
TokenChronicle Insights
Use only the user's configured TokenChronicle data directory.
Resolve the plugin root and use its exact matching signed executable under
bin/<platform>/tokenchronicle/ when
present; never download or substitute a binary. If an installed plugin has no matching executable,
report that the platform client has not been published and stop. When a report refresh is needed, run
the matching executable's memory-daily command. An unpacked development bundle
may use its bundled Python package, but must select the host's Python launcher and environment syntax;
do not use a Unix PYTHONPATH=... python3 command unchanged on Windows.
Run the read-only readiness command first. If TokenChronicle is not initialized or the first archive
is incomplete, switch to guided setup instead of analyzing an empty default location. If daily
protection is inactive, report that limitation with the analysis date.
- Read derived reports before opening detailed conversation files.
- Distinguish observed facts from inferred categories.
- For work, life, domain, preference, or memory analysis, include source date, project scope, and confidence.
- Treat personal preferences as candidates until the user approves them.
- Do not modify Codex memories or upload source conversations.
想直接用这个技能?
本站把开放许可(MIT / Apache 等)的技能按仓库打包整理到网盘,点一下转存到你自己的网盘,不用一个个从 GitHub 拉。许可未声明的技能只给原始仓库链接,不打包。
它属于哪个仓库
plugins/HuaHuiJian/tokenchronicle-marketplace/skills/tokenchronicle-insights/SKILL.md