feat: CLI operations package, new ES agent skills, and SDK 0.1.1-alpha.2

.github/agents/applogger-delivery-engineer.agent.md

@@ -31,6 +31,8 @@ Load the relevant workflow skills under `.github/skills/` based on the task:
 8. Do not create a tag automatically after every merge to `main`.
 9. Documentation-only work may go `dev -> main` without any tag.
 10. Tag only when the merged change is release-worthy or the user explicitly wants a versioned release.
+11. Do not push every micro-change; batch work locally and push only when a meaningful milestone is complete (full feature slice, completed documentation unit, or explicit checkpoint requested by the user).
+12. If the change is docs-only, skip manual build/test validation and run only documentation-focused checks.
 
 ## Parallelization Strategy
 
@@ -47,15 +49,16 @@ Use parallel subagents only for exploration and audit. Do not parallelize writes
 1. Load repo context and identify impacted areas.
 2. If needed, launch parallel read-only subagent passes.
 3. Complete or review code and docs changes.
-4. Run local validation, including `act`-based local GitHub Actions when relevant.
-5. Ensure final delivery branch is `dev`.
-6. Push `dev` only after local validation passes.
-7. Open and verify PR from `dev` to `main`.
-8. Merge only after checks pass.
-9. Decide whether the merged change is tag-eligible.
-10. Only after `main` is verified and the change is tag-eligible, create and push the release tag.
-11. Review open Dependabot PRs and merge only when technically sound.
-12. Confirm docs and release state are current.
+4. Classify the change: code, docs-only, or mixed.
+5. Run local validation relevant to the changed area, including `act`-based local GitHub Actions only when workflow-sensitive.
+6. Ensure final delivery branch is `dev`.
+7. Push `dev` only after the current milestone is complete and validation passes.
+8. Open and verify PR from `dev` to `main`.
+9. Merge only after checks pass.
+10. Decide whether the merged change is tag-eligible.
+11. Only after `main` is verified and the change is tag-eligible, create and push the release tag.
+12. Review open Dependabot PRs and merge only when technically sound.
+13. Confirm docs and release state are current.
 
 ## Intent Interpretation
 
@@ -64,6 +67,7 @@ Interpret user requests strictly:
 1. `merge dev to main` means complete integration to `main`, not release tagging.
 2. `create tag`, `publish release`, or `ship version` means evaluate release-tag eligibility first.
 3. If the change is docs-only, workflow-only, or internal maintenance, merge may be correct while tagging is not.
+4. If the user asks to continue a larger feature (for example CLI construction), keep iterating locally and delay push/PR until the requested feature slice or milestone is complete.
 
 ## Output Format
 

.github/skills/applogger-change-delivery/SKILL.md

@@ -27,18 +27,21 @@ Trigger phrases include:
 5. Review docs impact before push.
 6. Do not assume every merged PR should become a release tag.
 7. Treat merge-to-main and tag creation as separate decisions.
+8. Do not push for every small edit; push when a meaningful milestone is complete (full feature slice, completed docs package, or explicit user checkpoint).
+9. For docs-only changes, skip manual build/test execution and run documentation-focused checks only.
 
 ## Procedure
 
 1. Read [completion checklist](./references/completion-checklist.md).
-2. Run [local validation](./references/local-validation.md).
-3. Run [local GitHub Actions validation](./references/local-actions-validation.md) when workflow-sensitive.
-4. Audit docs with [documentation gate](./references/documentation-gate.md).
-5. Commit only after the feature/fix/refactor is complete.
-6. Push the integrated change to `dev`.
-7. Open PR from `dev` to `main`.
-8. Merge only after checks are green.
-9. Decide whether the merged change is release-tag eligible before invoking the release flow.
+2. Classify the change as code, docs-only, or mixed.
+3. Run [local validation](./references/local-validation.md) using the profile that matches the change class.
+4. Run [local GitHub Actions validation](./references/local-actions-validation.md) only when workflow-sensitive.
+5. Audit docs with [documentation gate](./references/documentation-gate.md).
+6. Commit only after the feature/fix/refactor/docs milestone is complete.
+7. Push the integrated milestone to `dev`.
+8. Open PR from `dev` to `main`.
+9. Merge only after checks are green.
+10. Decide whether the merged change is release-tag eligible before invoking the release flow.
 
 ## Parallelization Guidance
 
@@ -51,6 +54,7 @@ For larger changes, run read-only subagent passes in parallel before editing or
 ## Output Standard
 
 1. Report local validation results.
-2. Report whether docs changed or were intentionally unchanged.
-3. Report push/PR status.
-4. State whether the change is tag-eligible, not tag-eligible, or needs explicit release intent.
+2. Report the selected validation profile (code, docs-only, or mixed).
+3. Report whether docs changed or were intentionally unchanged.
+4. Report push/PR status.
+5. State whether the change is tag-eligible, not tag-eligible, or needs explicit release intent.

.github/skills/applogger-change-delivery/references/completion-checklist.md

@@ -11,4 +11,5 @@ Before delivery:
 Commit policy:
 
 1. Do not create the final delivery commit for partial work.
-2. Use one clear commit or a clean, intentional set of commits for the completed unit.
+2. Use one clear commit or a clean, intentional set of commits for the completed unit.
+3. Avoid micro-commits for every small step; commit and push when a milestone is complete.

.github/skills/applogger-change-delivery/references/local-validation.md

@@ -1,6 +1,10 @@
 # Local Validation
 
-Baseline local validation:
+Choose validation by change type.
+
+## Profile A: Code Changes (SDK or executable logic)
+
+Run baseline local validation:
 
 1. `cd sdk && ./gradlew check`
 2. `cd sdk && ./gradlew assemble`
@@ -10,4 +14,19 @@ Push-gate validation:
 1. `.githooks/pre-push` runs Detekt.
 2. `.githooks/pre-push` runs `:logger-core:jvmTest` and `:logger-test:jvmTest`.
 
-Use the exact pre-push hook when you need parity with local gatekeeping.
+Use the exact pre-push hook when you need parity with local gatekeeping.
+
+## Profile B: Docs-Only Changes
+
+When no executable code changed, skip manual build/test runs.
+
+Run only documentation-focused checks, for example:
+
+1. Verify modified docs render and links are valid.
+2. Verify examples and command snippets match current behavior.
+
+If a push still triggers repository hooks or CI checks, allow them to run naturally, but do not add extra manual build/test commands.
+
+## Profile C: Mixed Changes
+
+Apply Profile A for code validation plus docs checks from Profile B.

.github/skills/applogger-cli-agent-operator/SKILL.md

@@ -0,0 +1,67 @@
+---
+name: applogger-cli-agent-operator
+description: 'Operate the AppLoggers CLI as a deterministic interface for AI agents and automation. Use this when an agent must query CLI metadata, capabilities, health, and telemetry output in machine-readable form.'
+argument-hint: 'Describe what automation or agent task should be executed through applogger-cli and which output contract is required.'
+user-invocable: true
+---
+
+# AppLogger CLI Agent Operator
+
+## Purpose
+
+Use this skill when a Copilot agent or automation needs to interact with AppLoggers through the CLI contract in a stable and parseable way.
+
+## Hard Rules
+
+1. Prefer `--output agent` for machine consumption (TOON compact encoding via toon-go).
+2. Use `--output json` when downstream systems require JSON strictly.
+2. Treat exit code 0 as success, 1 as runtime failure, 2 as usage error.
+3. Run `agent schema` and `capabilities` before implementing new automation assumptions.
+4. Do not parse free-form text when JSON is available.
+
+## Standard Command Set
+
+1. Metadata discovery:
+   - `applogger-cli --syncbin-metadata --output json`
+2. Version/build discovery:
+   - `applogger-cli version --output json`
+3. Capability discovery:
+   - `applogger-cli capabilities --output agent`
+4. Agent contract discovery:
+   - `applogger-cli agent schema --output agent`
+5. Runtime readiness probe:
+   - `applogger-cli health --output agent`
+6. Telemetry query endpoint:
+   - `applogger-cli telemetry query --output agent`
+7. Dedicated compact orchestration response:
+   - `applogger-cli telemetry agent-response --source logs --aggregate severity --preview-limit 5`
+
+## Supabase Environment Setup
+
+Before `telemetry query`, ensure environment is configured:
+
+1. Required:
+   - `APPLOGGER_SUPABASE_URL`
+   - `APPLOGGER_SUPABASE_KEY` (service_role key)
+2. Optional:
+   - `APPLOGGER_SUPABASE_SCHEMA`
+   - `APPLOGGER_SUPABASE_LOG_TABLE`
+   - `APPLOGGER_SUPABASE_METRIC_TABLE`
+   - `APPLOGGER_SUPABASE_TIMEOUT_SECONDS`
+3. If operating with Supabase MCP available, retrieve:
+   - project URL from `mcp_supabase_get_project_url`
+4. Provision `APPLOGGER_SUPABASE_KEY` from secure secret storage (service_role).
+   - Do not use publishable/anon keys for CLI read operations.
+
+## Error Handling Contract
+
+1. If exit code is 2, the caller should correct arguments and retry.
+2. If exit code is 1, treat as runtime error and escalate with captured stderr JSON envelope.
+3. Always store command, args, exit code, and full JSON response in agent logs for traceability.
+
+## Output Standard
+
+1. Report executed commands.
+2. Report parsed JSON fields used for decisions.
+3. Report retries and final status.
+4. Report remaining uncertainty if command is preview status.

.github/skills/syncbin-child-cli-architect/SKILL.md

@@ -10,6 +10,14 @@
 ## Purpose
 This skill enables AI agents to architect and build COMPLETE, production-ready Child CLIs for the Syncbin ecosystem from scratch. The agent will master the full lifecycle: requirements analysis, architecture design, implementation following Syncbin standards, testing, documentation, OCI packaging, and marketplace publishing. This is the comprehensive skill for creating professional-grade CLI tools.
 
+## AppLoggers Delivery Cadence
+
+When this skill is used inside the AppLoggers repository:
+
+1. Do not push every incremental step while building the CLI.
+2. Push only after a complete CLI milestone is closed (for example: foundation complete, telemetry query module complete, installer module complete).
+3. If the milestone is docs-only, do not run manual build/test commands before push.
+
 ## Core Concepts
 
 ### 1. The Syncbin Child CLI Standard