Personal Skills 0.33.1

Personal Skills 0.33.1

• session-meter 0.2.1: the band shows the time until the five-hour plan window resets (5h 25% 2h14m) and the cache countdown drops its / 1h suffix.
• explore-model README: which model the haiku alias resolves to on each provider, and how to pin one.

Personal Skills 0.33.0

• explore-model 0.1.0, a Claude Code mod in the Claude catalog. The
  built-in Explore subagent runs on the main conversation's model unless the
  Agent call names one; this fills model and effort (default Haiku,
  medium) on Explore calls that name neither. Explore keeps Anthropic's prompt
  and read-only tools, and a call that names its own model or effort keeps it.
• benchmarks/explore-model/, the measurement behind it: three rounds on
  audiobook-boss in which Haiku matched the main-model Explore at about 1/30
  the cost. Not a plugin; no catalog lists it.

Personal Skills 0.32.0

• done-gate 0.1.0, a Claude Code mod in the Claude catalog. It blocks the
  main loop from ending a turn while changed files have not passed the
  repository's check: a Bash command matching checks in
  .claude/done-gate.json that ends without error records a git tree id of
  the whole working tree, and a stop with files changed since then blocks
  once, then ends with a toast. A resolve command lets the repository say
  which checks are unmet (audiobook-boss answers from its verify.sh lane
  map). Inactive without the config file.

• cache-meter is now session-meter 0.2.0. The band adds context fill and
  five-hour and weekly plan-limit use; /meter (was /cache) adds limit
  bars with reset times and one row per subagent with its turns, tokens and
  share of the session's tokens. Reinstall as session-meter@personal;
  options saved under cache-meter reset to defaults.

• issue-hygiene-reminder 0.1.1: its Codex manifest passes the pinned Codex
  validator. The hooks field is gone (Codex loads the default
  hooks/hooks.json, the same file) and an interface block is added.

• oura-mcp 1.1.1: the public hostname is accepted only when
  MCP_ACCESS_TOKEN is set, and an empty token variable counts as unset.

Personal Skills 0.31.0

Personal Skills 0.31.0

Personal Skills 0.30.0

Personal Skills 0.30.0

Personal Skills 0.29.1

Personal Skills 0.29.1

Fixed: nested-agents-md 0.1.1 also loads nested AGENTS.md for a path after cd in the same command, and for a path spelled with ..

Personal Skills 0.29.0

Personal Skills 0.29.0

Added nested-agents-md 0.1.0, a Claude Code mod. Claude Code loads a nested AGENTS.md only when Claude reads a file with the Read tool. This mod also loads it when a Bash command names a file, or Edit or Write changes one (anthropics/claude-code#90450). Claude only.

Personal Skills 0.28.0

Personal Skills 0.28.0

• All 19 personal-skills descriptions are rewritten. Each one the agent can
  load says what the skill does and when to use it; each user-only one says
  what you get when you call it.
• task-compass, improve-codebase-architecture, and security-best-practices
  are user-only in every client (Claude, Cursor, Grok, Codex). Call them by
  name. pre-pr-gut-check reads the last two from their files.

Personal Skills 0.27.2

complexity-lens 0.1.4 starts its list over with each new message and colours rows by how much they need attention: red when this request added a function over the limit, yellow when the file has one, dim otherwise.

Personal Skills 0.27.1

complexity-lens 0.1.3 and whittle's complexity ranker measure each file inside its own repository, so files outside the session's repository no longer fail with rank exit 1.