chore(ai-tutors): regenerate lesson-08 knowledge base

[potiuk]

Summary

• inject-knowledge-base.py --check reported ai-tutors/lesson-08-eval-driven-development.md as out of date on main. This is the regeneration; --check is now clean across all eleven prompts.
• The drift is real, not cosmetic: the tutor prompt embeds its source page verbatim, and docs(education): align eval worked example 2 with case-7 fixture #985 replaced worked example 2 with the actual case-7 fixture — a BashOperator KeyError report whose body opens with a SYSTEM: injection. The prompt still described a dark-mode feature request carrying a different injection string. A tutor teaching prompt-injection defence from an example that no longer exists in the fixtures is the wrong thing to hand a learner.
• Found while closing out fix(ai-tutors): handle nested code fences #1010 / fix(ai-tutors): stop nested code fences inverting the fence state #1011 (nested code fences). Nothing in this diff is fence-related — I verified both lesson-04 and lesson-08 were already stale on main before that work, so this is independent drift and was deliberately kept out of fix(ai-tutors): handle nested code fences #1010 rather than bundled in.

Type of change

• Skill change (.claude/skills/<name>/) — eval fixtures updated below
• Tool / bridge contract (tools/<system>/*.md)
• Python package (tools/*/ with pyproject.toml)
• Groovy reference impl
• Cross-cutting (RFC, AGENTS.md, sandbox, privacy-LLM)
• Documentation (docs/, README.md, CONTRIBUTING.md) — generated tutor prompt
• Project template (projects/_template/)
• CI / dev loop (prek, workflows, validators)
• Other: regenerated artefact — python3 ai-tutors/inject-knowledge-base.py output, no hand edits

Test plan

• prek run passes on the changed file (markdownlint, lychee, typos, SPDX, doctoc)
• python3 ai-tutors/inject-knowledge-base.py --check exits clean — before this change it listed lesson-08
• For Python packages touched — none
• For Groovy bridges touched — n/a
• For skill changes — no skill changed
• Other: confirmed the diff contains no code-fence changes, i.e. this is a content refresh rather than an artefact of the recent tag_bare_code_fences fix

RFC-AI-0004 compliance

• HITL — any new mutation is gated on explicit user confirmation
• Sandbox — no new unrestricted host access; network reach declared in the adapter
• Vendor neutrality — placeholders used in all skill / tool prose
• Conversational + correctable — agentic-override path documented if behaviour is adopter-tunable
• Write-access discipline — no autonomous outbound messages; drafts only, sent on confirmation
• Privacy LLM — private content does not reach a non-approved LLM; redactor invoked where needed

(None apply — regenerated documentation content.)

Linked issues

Refs #985 (the source-page change this propagates). Follow-up to #1010, which closed the related lesson-04 drift.

Notes for reviewers (optional)

The whole diff is generator output, so the useful review question is whether the generator should have been run, not whether the text is right. Two things worth knowing:

1. Nothing enforces this. No CI job or prek hook runs inject-knowledge-base.py --check, so a source-page edit can silently leave a tutor prompt stale — which is exactly how this one drifted. Now that ai-tutors/ is a uv workspace member (added in fix(ai-tutors): stop nested code fences inverting the fence state #1011), a --check step would be cheap to wire in. Worth a follow-up issue rather than folding into this PR.
2. chore(ai-tutors): regenerate knowledge base for lesson-04-your-first-skill #928 was closed as superseded rather than merged: it regenerated lesson-04 before two source-page changes landed, so merging it would have rolled that file backwards. This PR is the remaining half of the same cleanup.