v1.1.0 — 2026 prior art and audit gaps closed

v1.1.0 — 2026 prior art and audit gaps closed

Cumulative release of the agent-native-cli skill incorporating three
revision passes grounded in 2025–2026 writing on agent-native CLIs
(Anthropic, Enyioha, openstatus, Zechner, Ronacher, gh CLI, clig.dev,
sysexits.h).

Highlights since v1.0.0:

Rubric: 10 → 14 criteria with full principle coverage. Added
Three-audience support (P0), Boundary validation (P5), Idempotent
retries (P1), Non-interactive operation (P0). Scoring rebased to max 28.

New principles content:

• P1: TTY detection as the human/machine switch.
• P3: progressive / lazy schema disclosure (Anthropic code-execution-MCP).
• P4: sandboxing footnote (Anthropic "Beyond permission prompts").
• P6: schema versioning, deprecation, replaced_by signals.

New sections:

• Reducing agent round-trips (six concrete tactics).
• When CLI is the right answer (and when it isn't) — CLI vs MCP vs hybrid.
• References — 9 cited primary sources.

New design guidance:

• Idempotency and retry (--idempotency-key).
• Non-interactive operation.
• Long-running and streaming output.
• Partial-success envelopes.
• Optional meta telemetry slot.
• Locale, time, and determinism.

New worked examples:

• Example 6: idempotent batch with partial success, next hint, meta.
• Example 7: long-running export with structured stderr progress.

Fixes:

• Example 3 no longer teaches the agent to invoke auth token itself.
• Frontmatter trimmed 1103 → 906 bytes (under 1024 portability limit).
• Description is trigger-only with no feature list.
• README topics list aligned with live GitHub repo (18 topics).
• agent-delegated quick install instructions added to both READMEs.

v1.0.1

Changes

• Fix pi-mono naming and add correct repo link (https://github.com/badlogic/pi-mono)
• Update Codex install paths (~/.agents/ → ~/.codex/)
• Add MIT LICENSE file
• Add scripts/validate-metadata.sh for metadata consistency checks
• Clarify sidecar metadata language in docs

Full Changelog: v1.0.0...v1.0.1

v1.0.0 — Initial Release

agent-native-cli v1.0.0

First release of the Agent-Native CLI design & review skill.

What's included

• SKILL.md — Full skill with 7 principles, interaction loop, examples, non-examples, 10-criterion output rubric, and review checklist
• Multi-platform support — Claude Code, OpenClaw/ClawHub, Hermes Agent, Pi-Mo, OpenAI Codex, SkillsMP
• agents/openai.yaml — OpenAI Codex sidecar configuration
• README.md / README_CN.md — English and Chinese documentation

Core capabilities

• Evaluate CLI agent-readiness across 7 design principles
• Design stdout JSON contracts with stable error envelopes
• Define safety tiers: open / warned / hidden
• Design schema-driven self-description and dry-run previews
• Design delegated authentication with directional trust model
• Score CLIs on a 10-criterion rubric (0–2 per criterion)
• Produce P0/P1/P2 prioritized refactor plans