본문으로 건너뛰기
cccskills
무료GitHub에 공개

amp-continue: Thread-based conversation continuation with GF(3)

Combines AMP thread export with GF(3)-classified continuation for skill orchestration. Export threads as markdown, classify continuation choices as {-1:reductive, 0:neutral, +1:generative}, and fork parallel execution paths.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

설치 방법 보기

포함된 파일(3)

  • SKILL.md5.7 KB
  • CONCOMITANT_SKILLS.md1.5 KB
  • skill.py15.1 KB

SKILL.md(원문)

설치하기 전에 에이전트에게 전달되는 지시 내용을 확인할 수 있습니다.

Overview

The amp-continue skill bridges the AMP thread management system with Claude Code skill continuation. It:

  1. Exports AMP threads to structured markdown
  2. Classifies continuation choices using GF(3) ternary logic
  3. Forks parallel skill execution paths
  4. Preserves thread lineage through ACSet morphisms

GF(3) Continuation Matrix

ChoicePolaritySemanticsEffect
-1MINUSReductive continuationPrune search space, apply filters, narrow scope
0ZERONeutral continuationContinue linearly, preserve state, no branching
+1PLUSGenerative continuationExpand search space, generate variants, fork paths

Skill Interface

amp.export

Export threads to markdown with maximum detail.

(amp/export {:threads [thread-ids]
             :format :markdown
             :include-metadata true
             :include-messages true})

Output:

  • Markdown file with thread content
  • Metadata JSON
  • Message count and timestamps

amp.continue

Classify and fork continuation.

(amp/continue {:thread-id thread-id
               :decision :plus   ; or :minus, :zero
               :skills [skill-names]
               :parallel true})

Decision Classification:

  • :plus → Expand to N parallel execution paths
  • :zero → Continue serially to next skill
  • :minus → Apply filters to reduce paths

Implementation Details

SplitMix64 Determinism

def continue_path(thread_id: str, seed: int, choice: int) -> int:
    """
    Deterministic path selection.
    Same thread_id + same seed → same path always.
    """
    x = hash_thread_id(thread_id) ^ seed
    x = splitmix64_next(x)
    return x % num_paths

ACSet Morphism

Thread metadata preserved as ACSet:

threads : Thread → {id, title, created, updated, messages}
continuations : Continuation → {source_thread, choice, target_skill, seed}

Morphism: Thread → Continuation → Skill

Parallelism Pattern (Work-Stealing)

@dataclass
class ContinuationTask:
    thread_id: str
    skill_name: str
    decision: int  # GF(3): -1, 0, +1
    seed: int

def execute_continuation(task: ContinuationTask):
    # Steal from global task queue
    # Fork if decision == +1 (PLUS)
    # Continue if decision == 0 (ZERO)
    # Filter if decision == -1 (MINUS)

Examples

Export All Recent Threads

amp continue export \
  --threads "T-019b5df7..." "T-019b5df0..." \
  --output threads.md \
  --max-depth 3

Fork Generative Continuation

amp continue fork \
  --thread "T-019b5df7-50e8-744e-a95a-eb8d3fd72927" \
  --decision +1 \
  --skills "skill-maker" "chromatic-walk" "cq-ai" \
  --parallel 8

Apply Reductive Filter

amp continue filter \
  --thread "T-019b5da5-ee78-7113-b839-5c066cc98d4a" \
  --decision -1 \
  --selector "messages.length > 10" \
  --target-skill "summarize"

Continue Linear

amp continue next \
  --thread "T-019b5ddc-ea96-72ff-b26a-270b7e4527a1" \
  --decision 0 \
  --skill "load-sicp"

API Reference

(amp-continue/export opts)

Export threads with markdown generation.

Options:

  • :threads - Vector of thread IDs
  • :format - :markdown (default), :json, :html
  • :include-metadata - Boolean (default: true)
  • :include-messages - Boolean (default: true)
  • :output-file - String path (optional)

Returns: Markdown string + output file path

(amp-continue/continue opts)

Fork continuation based on GF(3) decision.

Options:

  • :thread-id - String thread ID
  • :decision - Keyword :plus, :zero, :minus
  • :skills - Vector of skill names to continue with
  • :parallel - Boolean or integer (worker count)
  • :seed - Integer (optional, for reproducibility)

Returns: ACSet of continuation tasks

(amp-continue/classify opts)

Classify thread continuation strategy.

Options:

  • :thread-id - String
  • :heuristic - :message-count, :complexity, :entropy

Returns: Classified decision keyword

MCP Integration

Registered as MCP tool:

{
  "name": "amp_continue",
  "description": "Export AMP threads and fork GF(3)-classified continuation",
  "resources": [
    {"uri": "amp://threads", "name": "All threads", "type": "read"},
    {"uri": "amp://continuations", "name": "Active continuations", "type": "read"}
  ]
}

Storage

  • Threads: ~/.local/share/amp/threads/*.json
  • Continuations: ~/.local/share/amp/continuations/*.json
  • Exports: ~/.local/share/amp/exports/

Performance

  • Thread load: O(1) per thread (disk lookup)
  • Export markdown: O(n) where n = total messages
  • Continuation fork: O(1) with work-stealing queue
  • Parallelism: 8x speedup with 8 workers on GF(3) branching

Testing

# Test export
pytest tests/amp_continue/test_export.py -v

# Test GF(3) branching
pytest tests/amp_continue/test_gf3_branching.py -v

# Test parallelism
pytest tests/amp_continue/test_parallelism.py -v

Related Skills

  • skill-maker - Meta-skill that generates amp-continue instances
  • chromatic-walk - Color-based thread path visualization
  • cq-ai - Security analysis of thread content
  • triadic-skill-orchestrator - GF(3) skill composition

Author

Generated by skill-maker meta-skill via GF(3) ternary adaptation.

리뷰

아직 리뷰가 없습니다. 사용해 본 소감을 남겨 주세요.

댓글

로그인 없이 쓸 수 있습니다. 사용 후기, 질문, 작동하지 않은 환경 등을 편하게 남겨 주세요.

아직 댓글이 없습니다. 첫 댓글을 남겨 보세요.

보내면 바로 공개됩니다. 이메일 주소나 전화번호 같은 개인 정보는 쓰지 마세요. URL은 2개까지입니다.

0 / 1,000자

이 양식은 reCAPTCHA로 보호되며 Google 개인정보처리방침 및 서비스 약관이 적용됩니다.

같은 저장소의 스킬

Model ElevenLabs voice synthesis as an ACSet schema with Unison ability handlers for typed voice operations. Use when integrating ElevenLabs TTS into category-theoretic pipelines, building typed voice synthesis workflows, or managing multi-voice narration with GF(3) conservation.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

plurigrid/asi702026년 7월 10일 업데이트

Systematic process for analyzing and modernizing legacy applications to comply with the 12-Factor App methodology. Covers code analysis, configuration externalization, dependency management, and container/Kubernetes best practices.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

plurigrid/asi702026년 7월 10일 업데이트

13c0e3

무료

Gay MCP skill at invocation 26

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

plurigrid/asi702026년 7월 10일 업데이트

19d57d

무료

Gay MCP skill at invocation 5

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

plurigrid/asi702026년 7월 10일 업데이트

1ea91a

무료

Gay MCP skill at invocation 27

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

plurigrid/asi702026년 7월 10일 업데이트

1eab21

무료

Gay MCP skill at invocation 3

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

plurigrid/asi702026년 7월 10일 업데이트

plurigrid의 스킬 모두 보기

이 스킬의 문제 신고