session-lifecycle-patterns
Manage AI agent session lifecycles with structured phases (FRAME, SHAPE, BUILD, PROVE), context preservation across sessions, handoff protocols, and session metadata tracking. Triggers on session management, agent lifecycle, or multi-session workflow requests.
What this skill does
# Session Lifecycle Patterns
Structure AI agent sessions for predictable outcomes, context preservation, and clean handoffs.
## The Phase Model
```
FRAME → SHAPE → BUILD → PROVE → DONE
↑ ↑ ↑ ↑
└───────┴───────┴───────┘
(back-transitions allowed)
```
### Phase Definitions
| Phase | Purpose | Activities | Gate |
|-------|---------|-----------|------|
| **FRAME** | Understand context | Explore code, read docs, ask questions | Can articulate the problem |
| **SHAPE** | Design approach | Create plan, identify files, consider tradeoffs | Plan reviewed and approved |
| **BUILD** | Implement | Write code, create files, apply changes | Implementation complete |
| **PROVE** | Verify | Run tests, validate output, check quality | All checks pass |
| **DONE** | Close | Summarize, commit, document decisions | Session artifacts preserved |
### Phase Transitions
```python
VALID_TRANSITIONS = {
"FRAME": ["SHAPE"],
"SHAPE": ["FRAME", "BUILD"], # Can go back to reframe
"BUILD": ["SHAPE", "PROVE"], # Can go back to reshape
"PROVE": ["BUILD", "DONE"], # Can go back to fix
"DONE": [], # Terminal
}
def can_transition(current: str, target: str) -> bool:
return target in VALID_TRANSITIONS.get(current, [])
```
### Phase Anti-Patterns
- **Skipping FRAME** — Jumping to code without understanding the problem
- **SHAPE → BUILD without approval** — Building before alignment on approach
- **Staying in BUILD forever** — Not proving that the work is correct
- **PROVE → DONE without evidence** — Claiming completion without verification
## Session Metadata
### Session Record Structure
```yaml
session:
id: "sess_2026-03-20_a1b2c3"
started: "2026-03-20T10:58:00Z"
ended: "2026-03-20T12:30:00Z"
phase_history:
- phase: FRAME
entered: "2026-03-20T10:58:00Z"
duration_minutes: 15
- phase: SHAPE
entered: "2026-03-20T11:13:00Z"
duration_minutes: 20
- phase: BUILD
entered: "2026-03-20T11:33:00Z"
duration_minutes: 45
- phase: PROVE
entered: "2026-03-20T12:18:00Z"
duration_minutes: 12
scope:
organ: IV
repo: a-i--skills
task: "Create python-packaging-patterns skill"
artifacts:
files_created: [skills/development/python-packaging-patterns/SKILL.md]
files_modified: []
commits: ["abc123"]
decisions:
- "Chose hatchling as recommended build backend"
- "Included namespace package pattern for multi-repo use"
```
## Context Preservation
### Intra-Session Context
Maintain context as the session progresses through phases:
```python
class SessionContext:
def __init__(self):
self.discoveries: list[str] = [] # FRAME findings
self.plan: dict = {} # SHAPE output
self.changes: list[str] = [] # BUILD artifacts
self.evidence: list[str] = [] # PROVE results
def frame_discovery(self, finding: str):
self.discoveries.append(finding)
def shape_decision(self, key: str, value: str, rationale: str):
self.plan[key] = {"value": value, "rationale": rationale}
```
### Inter-Session Context
For multi-session work, preserve essential context at session end:
```markdown
## Session Close Summary
### What was accomplished
- Created 6 Wave 0 skills (python-packaging-patterns through vector-search-patterns)
### What remains
- 18 more Wave 0 skills in Batches 2-4
- Waves 1-4 pending
### Key decisions
- Using governance_norm_group: repo-hygiene for packaging/config skills
- Using organ_affinity: [all] for cross-cutting infrastructure skills
### Blockers
- None
### Next session should start by
- Reading this summary
- Continuing with Wave 0 Batch 2
```
## Session Types
### Exploration Session
```
FRAME (80%) → SHAPE (20%) → DONE
```
Heavy on reading, light on planning, no implementation. Produces understanding and a plan for a future BUILD session.
### Implementation Session
```
FRAME (10%) → SHAPE (15%) → BUILD (55%) → PROVE (20%) → DONE
```
Quick reorientation, then focused building and verification.
### Debug Session
```
FRAME (40%) → BUILD (30%) → PROVE (30%) → DONE
```
Heavy investigation, targeted fix, thorough verification.
### Review Session
```
FRAME (30%) → PROVE (70%) → DONE
```
Mostly reading and validating existing work.
## Multi-Session Continuity
### Session Chains
When work spans multiple sessions, each session references the chain:
```yaml
chain:
id: "skill-fortification-campaign"
session_index: 3
total_sessions_estimated: 15
previous_session: "sess_2026-03-20_wave0-batch1"
completed_items: ["A1", "A2", "A3", "A4", "A5", "A6"]
remaining_items: ["A7", "A8", "A13", "A14", "B1", "B2", ...]
```
### Handoff Protocol
At session end, produce a handoff document:
1. **State snapshot** — What exists now (files, branches, test status)
2. **Decision log** — Why choices were made (so next session doesn't re-litigate)
3. **Next actions** — Exactly what to do next, with no ambiguity
4. **Risk flags** — Anything the next session should watch for
## Session Quality Signals
| Signal | Healthy | Unhealthy |
|--------|---------|-----------|
| Phase transitions | Sequential with occasional back-steps | Skipping phases, staying in one phase |
| FRAME duration | 10-30% of session | <5% or >50% |
| PROVE evidence | Concrete test/validation output | "I think this works" |
| Context preserved | Summary written at DONE | Session ends abruptly |
| Scope creep | Stays within stated scope | Expanding mid-BUILD |
## Anti-Patterns
- **Infinite FRAME** — Analysis paralysis; at some point, start shaping
- **Skipping PROVE** — Every BUILD must be proved before DONE
- **No session close** — Always write a summary, even if brief
- **Context loss between sessions** — Handoff documents are mandatory for multi-session work
- **Phase theater** — Going through motions without substance in each phase
Related in AI Agents
skill-development
IncludedComprehensive meta-skill for creating, managing, validating, auditing, and distributing Claude Code skills and slash commands (unified in v2.1.3+). Provides skill templates, creation workflows, validation patterns, audit checklists, naming conventions, YAML frontmatter guidance, progressive disclosure examples, and best practices lookup. Use when creating new skills, validating existing skills, auditing skill quality, understanding skill architecture, needing skill templates, learning about YAML frontmatter requirements, progressive disclosure patterns, tool restrictions (allowed-tools), skill composition, skill naming conventions, troubleshooting skill activation issues, creating custom slash commands, configuring command frontmatter, using command arguments ($ARGUMENTS, $1, $2), bash execution in commands, file references in commands, command namespacing, plugin commands, MCP slash commands, Skill tool configuration, or deciding between skills vs slash commands. Delegates to docs-management skill for official documentation.
reprompter
IncludedTransform messy prompts into well-structured, effective prompts — single or multi-agent. Use when: "reprompt", "reprompt this", "clean up this prompt", "structure my prompt", rough text needing XML tags and best practices, "reprompter teams", "repromptception", "run with quality", "smart run", "smart agents", multi-agent tasks, audits, parallel work, anything going to agent teams. Don't use when: simple Q&A, pure chat, immediate execution-only tasks. See "Don't Use When" section for details. Outputs: Structured XML/Markdown prompt, quality score (before/after), optional team brief + per-agent sub-prompts, agent team output files. Success criteria: Single mode quality score ≥ 7/10; Repromptception per-agent prompt quality score 8+/10; all required sections present, actionable and specific.
adaptive-compaction
IncludedAdaptive add-on policy and recovery layer that decides WHEN to compact, prune, snapshot, or fork -- replacing fixed-percent auto-compaction across Claude Code, Codex, and MCP-capable hosts. Trigger on auto-compact timing or damage: "when should I compact", "is it safe to compact now or start a fresh session", "auto-compact fires too early/mid-task", "switching to an unrelated task but the window still has space", "context rot", "answers get worse the longer the session runs", "the agent forgot the plan or my decisions after it summarized", "add a layer on top that manages context without changing the agent", raising autoCompactWindow to give the policy room, or installing/tuning a cross-tool compaction policy or PreCompact hook -- even when "compaction" is never said but the problem is context-window pressure or post-summarization memory loss. Do NOT use to summarize a conversation, build RAG, write a summarization prompt (decides WHEN not HOW), or answer max-context-length trivia.
agent-skill-creator
IncludedCreate cross-platform agent skills from workflow descriptions. Activates when users ask to create an agent, automate a repetitive workflow, create a custom skill, or need advanced agent creation. Triggers on phrases like create agent for, automate workflow, create skill for, every day I have to, daily I need to, turn process into agent, need to automate, create a cross-platform skill, validate this skill, export this skill, migrate this skill. Supports single skills, multi-agent suites, transcript processing, template-based creation, interactive configuration, cross-platform export, and spec validation.
llm-wiki
IncludedUse when building or maintaining a persistent personal knowledge base (second brain) in Obsidian where an LLM incrementally ingests sources, updates entity/concept pages, maintains cross-references, and keeps a synthesis current. Triggers include "second brain", "Obsidian wiki", "personal knowledge management", "ingest this paper/article/book", "build a research wiki", "compound knowledge", "Memex", or whenever the user wants knowledge to accumulate across sessions instead of being re-derived by RAG on every query.
skill-master
IncludedAgent Skills authoring, evaluation, and optimization. Create, edit, validate, benchmark, and improve skills following the agentskills.io specification. Use when designing SKILL.md files, structuring skill folders (references, scripts, assets), ingesting external documentation into skills, running trigger evals, benchmarking skill quality, optimizing descriptions, or performing blind A/B comparisons. Keywords: agentskills.io, SKILL.md, skill authoring, eval, benchmark, trigger optimization.