AgentSkillsCN

rai-session-start

在回顾验证、合并、清理与进度更新之后，完成一次故事。在审查结束后使用此功能，以正式结束故事的生命周期。

SKILL.md
--- frontmatter
name: rai-session-start
description: >
  Begin a session by loading context bundle, interpreting it, and proposing work.
  CLI does all data plumbing; skill does inference interpretation.

license: MIT

metadata:
  raise.work_cycle: session
  raise.frequency: per-session
  raise.fase: "start"
  raise.prerequisites: ""
  raise.next: ""
  raise.gate: ""
  raise.adaptable: "true"
  raise.version: "4.0.0"

Session Start

Purpose

Load deterministic context bundle from CLI, interpret it, propose focused work. The CLI assembles all data (profile, session state, memory graph, deadlines, coaching). The skill only does inference: interpret and present.

Mastery Levels (ShuHaRi)

Experience level is in the context bundle — adapt output verbosity accordingly.

  • •Shu: Detailed explanations, teach concepts
  • •Ha: Balanced output, explain new concepts only
  • •Ri: Minimal output, essentials only

Steps (2)

Step 1: Load Context Bundle

bash
rai session start --project "$(pwd)" --context

This single command:

  • •Loads developer profile from ~/.rai/developer.yaml
  • •Loads session state from .raise/rai/session-state.yaml
  • •Queries memory graph for foundational patterns
  • •Assembles a ~150 token context bundle
  • •Records the session start (increments count, sets active session)
  • •Warns about orphaned sessions if detected

First-time user: If no profile exists, ask for name:

bash
rai session start --name "Name" --project "$(pwd)" --context

If graph unavailable: Run rai memory build first, then retry.

Step 2: Interpret & Present

With the context bundle from Step 1, use inference to:

  1. •

    Check signals:

    • •Deadline pressure (<3 days → focus critical path)
    • •Coaching corrections → reinforce behavioral primes
    • •Pending decisions or blockers → address first
  2. •

    Check parking lot: If dev/parking-lot.md exists, scan for stale items (>2 weeks).

  3. •

    Propose session focus based on:

    • •Pending items from previous session (highest priority)
    • •Current story/phase (continue where left off)
    • •Deadlines (urgency modulation)
  4. •

    Present (adapt to experience level from bundle):

Ri output:

code
## Session: YYYY-MM-DD

**Context:** [Epic] → [Story], [phase], N days to deadline
**Focus:** [goal]
**Signals:** [any, or "None"]

Go.

Shu output adds: explanation of context, progress metrics, concepts.

Output

ItemDestination
Session summaryDisplayed (not saved)
SignalsDisplayed
Session state~/.rai/developer.yaml (via CLI in Step 1)

Notes

  • •One CLI call does all data plumbing — no separate profile/memory/graph queries
  • •Context bundle is deterministic — same inputs produce same output
  • •Skill is a thin inference layer — interpret, don't gather
  • •Foundational patterns in bundle serve as behavioral primes

References

  • •Context bundle: rai session start --context
  • •Profile: ~/.rai/developer.yaml
  • •Session state: .raise/rai/session-state.yaml
  • •Memory graph: .raise/rai/memory/index.json
  • •Parking lot: dev/parking-lot.md
  • •Complement: /rai-session-close