RepoJournal

@anurag629

A short briefing every day anurag629 ships something — in about 3 minutes.

One email a day. Unsubscribe in one click.

361 stories across 44 days in 2026, 101-day streak

Mon Wed Fri
Mar Apr May Jun Jul Aug
an entry to read activity only Less More before entries began Full year →

Entries start Jan 13, 2022 — we reconstructed the previous month from @anurag629's public GitHub activity on the day they joined. Squares before that show contribution counts only.

Full archive →

Agent skills discovery and config file precedence

By RepoJournal , from @anurag629's public GitHub activity

anurag629
1 person shipped this
anurag629 @anurag629 5 cited

Anurag Verma added skill file discovery to the agent and rewrote config resolution to handle command-line overrides and environment variables.

In quantized-ai/luca-py, Verma shipped agent skills support [1], allowing the agent to load instruction sets from SKILL.md files on disk. Each skill file contains YAML frontmatter with a name and description, followed by a body that arrives only through a new skill tool. The implementation searches multiple locations in order [3]: workspace/.claude/skills, workspace/.agents/skills, configured extra_skill_locations, then global ~/.claude/skills and ~/.agents/skills.

Verma also completed the config file precedence work [2], adding both --config flag and LUCA_CONFIG_PATH environment variable support. A named config file replaces discovery entirely rather than adding another layer, so neither project nor home config interfere [5]. The upward walk for project-level luca.json was also bounded at the home directory [4], fixing a bug where a forgotten ~/.luca.json would silently apply to every non-repo directory under home.

One email a day. Unsubscribe in one click.

References

  1. [1] Agent Skills: read SKILL.md instruction sets from disk ↗ quantized-ai/luca-py
  2. [2] Global and local config: --config, LUCA_CONFIG_PATH, and project-scoped discovery ↗ quantized-ai/luca-py
  3. [3] Agent Skills: read SKILL.md instruction sets from disk (#20) quantized-ai/luca-py
  4. [4] Bound the project-config walk at the home directory (#21) quantized-ai/luca-py
  5. [5] Merge pull request #29 from quantized-ai/config-file-override quantized-ai/luca-py

Quick answers

What shipped in anurag629 on August 2, 2026?
Anurag Verma added skill file discovery to the agent and rewrote config resolution to handle command-line overrides and environment variables. In total, 3 commits and 2 pull requests landed.
Who contributed to anurag629 on August 2, 2026?
1 developer shipped this update, including anurag629.
What were the notable anurag629 updates?
Agent Skills: read SKILL.md instruction sets from disk, Global and local config: --config, LUCA_CONFIG_PATH, and project-scoped discovery, and Agent Skills: read SKILL.md instruction sets from disk (#20).

More from @anurag629's journal

@anurag629

Model catalog switched to models.dev

Anurag Verma replaced two drifting hand-maintained lists with a live feed from models.dev, fixing pricing data that was off by 3x for some models.

anurag629
2 shipped
@anurag629

Session storage overhaul and per-model system prompts

Anurag Verma landed two major features in quantized-ai/luca-py, moving session files out of project directories and adding model-aware prompts to the agent.

anurag629
4 shipped
@anurag629

Added Quantized as a built-in provider to luca-py

Anurag Verma integrated Quantized as an OpenAI-compatible provider and made progress on native subagent rendering in the TUI.

anurag629
4 shipped
@anurag629

OpenAI responses transport and response format handling

Anurag merged a major refactor of OpenAI integration in luca-py, moving to the /v1/responses endpoint and making response_format a real, opaque data structure that survives model switches.

anurag629
3 shipped

Follow along

Follow anurag629

A short briefing every day anurag629 ships something — in about 3 minutes.

or

One email a day. Unsubscribe in one click. Read a past issue →