chore(mule-development): add generate-connectivity-knowledge skill, bump to 1.0.5

[matiasjlopez]

Summary

• Add generate-connectivity-knowledge — a new 14-step skill that produces connectivity knowledge for a SaaS API when no dedicated Mule connector exists. Researches the API from user-defined use cases and documentation, produces an OpenAPI 3.0 spec, validates every operation against the live service with auto-fix, and writes a self-contained connectivity-schema/<apiName>/ folder (api-reference.md + <apiName>.yaml + config.properties).
• The output feeds the HTTP-fallback branch of build-mule-integration (Step 3a) so HTTP-Connector flows inherit the same auth, pagination, and entity awareness a dedicated connector would carry.
• Bump @salesforce/mulesoft-vibes-skills from 1.0.4 → 1.0.5 and add the corresponding CHANGELOG.md entry.

What's in the skill

File	Purpose
generate-connectivity-knowledge/SKILL.md	The 14-step workflow: research → OpenAPI 3.0 generation → live validation with auto-fix.
generate-connectivity-knowledge/references/examples.md	Worked examples for the most common patterns the agent encounters during research and validation.

The skill includes MANDATORY credential-security rules: never read, echo, log, or reason about credential values from config.properties. Verification is via grep -c '^KEY=' and injection via inline shell substitution; the values never enter the agent's transcript.

Test plan

• make pre-commit-hook — validate-descriptions, validate-mcp-server, validate-xorigin, validate-jtbd PASS
• make test-portal — pytest + jest PASS
• make validate-all-governed — 35 passed, 0 failed, 1 skipped (arm-monitoring-query, expected)
• python3 scripts/build/validate_jtbd.py skills/mule-development/generate-connectivity-knowledge/SKILL.md . — PASSED
• Local agent-evaluation smoke set authored (3 tasks: happy path, input gathering, credential security) on a clone of mule-dx-agent-evaluation; validate_instructions.py + eval_engine validate PASS for all three. Live execution against the bundled extension is blocked by a port-wiring mismatch between agentic-chat-1.0.17 (hardcoded PORT = 9876) and the eval runner (writes dynamic TEST_SERVER_PORT); handoff to Berkay Polat / Yanqi Luo for execution against their build is the documented next step.

Note on SKIP_PRE_PUSH: the pre-push hook intermittently fails on the first API in validate-all-governed because Anypoint CLI's plugin loader misbehaves when git pipes stdin to the hook. Standalone make validate-all-governed passes 35/35. CI will re-run all gates.

Out of scope

• Wiring build-mule-integration Step 3a to invoke the new skill — ships in a follow-up PR.
• Upstream PR of the 3 eval tasks to mule-dx-agent-evaluation — defer until Berkay/Yanqi review.

GUS

@W-22567718