Skip to content

Clarify improve evidence selection and execution context - #16

Merged
harshsinha03 merged 2 commits into
mainfrom
ab/improve-render-pr19
Oct 30, 2026
Merged

harshsinha03 merged 2 commits into
mainfrom
ab/improve-render-pr19

Conversation

@ashpreetbedi

@ashpreetbedi ashpreetbedi commented Oct 30, 2026 •

Copy link
Copy Markdown
Contributor

Bare /improve now selects simulations and local evaluations. Production evidence requires an explicit request and the User JWT from OS Security. The skill also clarifies ambiguous targets, verifies development execution destinations, and documents non-paginated session runs with time filters or single-run retrieval.

Port the complete skill from Render #19, pinned at 40f469f2cb04b1a35502f8c0240243390bb0ace2, reconciled through Railway. Only .agents/skills/improve/SKILL.md changes; the template attribution URL and existing .claude/skills symlink are preserved.

Validation: skill links (6 per template), exact attribution, preserved symlink, one-file scope, normalized content equality, whitespace and offline Agno 3.1.12 client signatures pass. Local serial gate passed: API build/start, /docs HTTP 200, MCP 9 tools, formatting/Ruff/mypy, documentation links and bounded smoke 3/3. Isolated containers and database volume were removed.

CI: all checks pass at bf7a8173c396a885a8e647b4769e6e2428910e6e.

Unrelated baseline drift is recorded separately and left unchanged. No production AgentOS API calls, cloud deployment, or simulation campaign. Local checks do not establish that a coding agent will always obey the skill; production access and hosted behavior are untested.

@harshsinha03
harshsinha03 merged commit 0f29d2c into main Oct 30, 2026
2 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants