菜单
Use when the user wants to view, search, add, edit, prune, archive, or clear hyperflow memory entries. CRUD interface for `.hyperflow/memory/` — never modifies source code, only memory files. Trigger with /hyperflow:cache, "show memory", "search memory for X", "clear memory", "what does hyperflow remember about Y".
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/TuYv/ccpm --skill cache命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
基于 SOC 职业分类
| name | cache |
| description | Use when the user wants to view, search, add, edit, prune, archive, or clear hyperflow memory entries. CRUD interface for `.hyperflow/memory/` — never modifies source code, only memory files. Trigger with /hyperflow:cache, "show memory", "search memory for X", "clear memory", "what does hyperflow remember about Y". |
| allowed-tools | Read, Write, Edit, Bash(ls:*), Bash(mv:*), Bash(rm:*), Glob, Grep, Agent, AskUserQuestion |
| argument-hint | <show|search|add|edit|prune|archive|clear|stats|migrate|off|compact> [args] |
| version | 3.1.3 |
| license | MIT |
| compatibility | Designed for Claude Code |
| tags | ["memory","persistence","project-state"] |
CRUD interface for .hyperflow/memory/. Full protocol: memory-system.md.
All operations target .hyperflow/memory/ at the project root. Never modify source code files — if asked to "remember X about file Y", add a memory entry only, never edit Y.
| Subcommand | Description |
|---|---|
show [tag] | Print index or filter entries by tag |
search <query> | Full-text search across all memory files |
add <category> <title> | Append a new entry (prompts for details) |
edit <entry-id> | Find entry by date+title slug and update in place |
prune | Remove stale, superseded, and orphaned entries |
archive | Move entries older than 30 days to cold storage |
clear | Wipe all memory (with confirmation, recoverable) |
stats | Counts, tier breakdown, tag frequency, oldest/newest |
migrate | Import entries from legacy ~/.claude/hyperflow-memory.md |
off | Disable memory writes for this session |
compact | Summarise aged memory entries into stubs + monthly archive sidecars |
show [tag]No arg → print index.md. With tag → filter all files for matching entries.
Output table: Date | Title | Tags | File | Tier
search <query>grep/ripgrep across learnings.md, decisions.md, pitfalls.md, patterns.md, conventions.md.
Return file:line + snippet, ranked by relevance.
add <category> <title>Categories: learning decision pitfall pattern convention
Prompt via AskUserQuestion for: what, why it matters, tags (controlled vocab).
Append to the matching file using:
### [YYYY-MM-DD] <title> `[tag1, tag2]`
**What:** ...
**Why it matters:** ...
**Evidence:** ...
That is the whole write. index.md is derived — scripts/memory-index.py rebuilds it at the next session start. Never append index rows by hand; to refresh it now, run python3 scripts/memory-index.py .hyperflow.
edit <entry-id>Locate by date+title slug. Show current value, prompt for new value, update in place.
prunePer memory-system.md pruning protocol:
[SUPERSEDED] entries older than 7 daystest -f).hyperflow/memory/archive/YYYY-MM.md
Print summary of removed/archived counts.archiveCompress hot entries older than 30 days → .hyperflow/memory/archive/YYYY-MM.md.
Leave one-line summary in original file. The stub keeps its date, so the derived index re-tiers it to cold on its own.
clearConfirm via AskUserQuestion: "This wipes all memory for this project. Are you sure?"
If yes → move all content to .hyperflow/memory/archive/cleared-<timestamp>.md, then reset files to empty stubs.
statsPrint: total entries, hot/warm/cold counts, tag frequency table, oldest and newest entry dates.
migrateRead ~/.claude/hyperflow-memory.md, filter entries matching current project path.
Append matching entries to learnings.md. Leave legacy file untouched.
Print count of migrated entries.
offPrint: "Memory writes disabled for this session." No files modified.
compactUser-invoked memory compaction. Summarises entries older than 7 days into stub lines and preserves the full text in monthly archive sidecars at .hyperflow/memory/archive/YYYY-MM.md.
Flow:
learnings.md; pass a path to target another).[domain, type] and legacy backticked `[domain, type]` tag forms are accepted.### [YYYY-MM-DD] Short title [domain, type] — summarized, see archive/YYYY-MM.md.archive/YYYY-MM.md, grouped by each entry's calendar month..hyperflow/memory/.checksums (a memory-scoped sidecar — distinct from .hyperflow/.checksums which the scaffold staleness check owns) and exits with a summary.Output: N entries compacted into archive/YYYY-MM.md · M stubs rejected as duplicates · source N→M lines. Full protocol in compaction.md.
/hyperflow:cache is the operator interface to project-scoped memory under .hyperflow/memory/. It's the only skill that mutates memory files directly (other skills append via the memory-system protocol). Subcommands cover the full lifecycle: show, search, add, edit, prune, archive, clear, stats, migrate. All operations are project-local — entries never leak across projects.
.hyperflow/ initialized (run /hyperflow:scaffold if missing — cache creates .hyperflow/memory/ on first write but expects the parent dir)..hyperflow/memory/ and .hyperflow/memory/archive/.migrate only: read access to ~/.claude/hyperflow-memory.md (legacy global memory).See Subcommands and Subcommand Details above for the full operational spec. Summary:
clear requires AskUserQuestion confirmation; migrate requires legacy file presence)..hyperflow/memory/.Each subcommand prints a compact summary:
show — table of matching entries (Date | Title | Tags | File | Tier).search — file:line matches with snippets, ranked by relevance.add / edit — confirmation line with new entry id and target file.prune / archive / clear — counts of removed/archived/cleared entries plus destination paths.stats — totals + hot/warm/cold breakdown + top-N tags.migrate — count of migrated entries + source legacy file path.off — single-line Memory writes disabled for this session.| Failure | Behavior |
|---|---|
.hyperflow/memory/ missing | Auto-create skeleton (index.md + 5 category files + archive/.gitkeep) on first write; for read-only subcommands, print (no memory yet — invoke /hyperflow:scaffold first). |
| Subcommand unknown | Print subcommands table; suggest closest match via Levenshtein distance. |
add with invalid category | Reject and list valid categories: learning, decision, pitfall, pattern, convention. |
edit entry id not found | List 3 closest matches by title slug + date. |
clear without confirmation (headless) | Refuse and print clear requires interactive confirmation. Do not wipe. |
migrate source file missing | Print (nothing to migrate — ~/.claude/hyperflow-memory.md not found) and stop. |
/hyperflow:cache show
Date Title Tags File Tier
2026-05-16 Bash scoping required by validator [validator, marketplace] learnings.md hot
2026-05-15 No AI attribution in commits [convention, git] conventions.md hot
2026-05-14 Per-task commits in plugin dev [convention, git] conventions.md hot
3 entries (3 hot, 0 warm, 0 cold)
/hyperflow:cache search "validator"
.hyperflow/memory/learnings.md:42 — "Jeremy's validator requires scoped Bash..."
.hyperflow/memory/decisions.md:8 — "...validator score of 73 → 94 after fix"
2 matches
/hyperflow:cache add learning "Markdown frontmatter needs block scalar for colons"
? What: Block scalar (|) preserves : and backticks in YAML values
? Why it matters: prevents fatal YAML parse failures in marketplace validators
? Tags: yaml, validator, frontmatter
Added — .hyperflow/memory/learnings.md (entry 2026-05-16-block-scalar-frontmatter)
/hyperflow:cache stats
Memory entries: 47
Hot (≤7d) 12
Warm (8-30d) 23
Cold (30d+) 12
Top tags: validator (8), convention (7), git (6), yaml (4)
Oldest: 2026-02-14 Newest: 2026-05-16
/hyperflow:cache compact protocol: stub format, archive sidecar, idempotency.