Two tools with reasonable naming (verb_noun pattern) but significant gaps in schema documentation and parameter descriptions. dump_session has a well-structured input schema with 7 parameters (6 optional), but get_context lacks visible input schema details beyond a single optional 'surface' enum parameter. Both descriptions are adequate (100-150 chars) but lack actionable guidance on when to use each tool vs. alternatives. No output schemas documented. No error handling guidance. Parameter descriptions are present but generic, e.g., 'Project name (matches your folder name in ~/projects/)' lacks format constraints or validation rules.

• Document output schema for both tools. For dump_session: return {success: boolean, session_id: string, updated_at: ISO8601, context_card_url?: string}. For get_context: return {projects: [{name, last_session, tags}], global_preferences: {...}, active_projects_count: number}.
• Add format constraints to dump_session parameters: project name must match /^[a-zA-Z0-9_-]+$/, tags must be lowercase alphanumeric + hyphens, max 50 chars per tag, max 10 tags total.
• Clarify idempotency: 'Calling dump_session twice with the same project and summary updates the existing session record (idempotent). Use unique summaries to create separate session entries.'
• Add error recovery guidance: 'If project not found, call get_context first to list available projects. If DB is locked, retry after 5 seconds.'
• Expand get_context description to explain when to use it vs. get_orientation: 'Use get_context for cross-project snapshots or when the context card was not injected. For single-project orientation, prefer get_orientation (faster, more focused).'

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry