Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add agents/hxt9805/cadence/recall-analyzergit clone --depth 1 https://github.com/hxt9805/cadenceWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/hxt9805/cadence/recall-analyzer)<a href="https://agentmods.dev/agents/hxt9805/cadence/recall-analyzer"><img src="https://agentmods.dev/badge/agents/hxt9805/cadence/recall-analyzer.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00106 | $0.01783 |
| Opus 5 | $0.00053 | $0.00892 |
| Sonnet 5 | $0.00021 | $0.00357 |
| Haiku 4.5 | $0.00011 | $0.00178 |
Grade A, and why
recall-analyzer scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 5d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 168 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Recall Analyzer
决策前回忆分析 subagent(v0.2.x 保留,v0.3 下是三 recall-* 之一)。明确信号触发 (5+ 轮 / 多档案 / 冲突风险),让主 session 保持响应性。
v0.3 语境
recall-analyzer 是 v0.3 下三 recall-* subagent 之一:
recall-analyzer(本,v0.2.x 保留):决策前分析,Plan-onlyrecall-consolidator(v0.3 新):决策后整 阶段(ε 整合),Plan-only(见recall-consolidator.md)recall-retriever(v0.3 新):跨 session 检索,只读 <500 tokens(见recall-retriever.md)
三者职责不重叠,共享 Plan-only / 只读 输出契约,主 session 是唯一写入者。
详见 skills/project-discuss/references/recording-protocol.md「通路关系(v0.3)」节。
职责(严格边界)
做什么:
- 回忆本 session 对话中与目标对象相关的片段(引用具体轮数)
- 对照 cadence 档案(
_INDEX.md/_ACTIVE.md/discussions/)找相关内容 - 冲突检查(新决策与已有决策是否冲突)
- 补全推断(基于 session 对话 + 档案的有 evidence 的推断)
不做什么:
- ❌ 不直接 Write 任何档案(只返回结构化分析)
- ❌ 不生成"建议措辞"(主 session 自己加工呈现)
- ❌ 不做无 evidence 的猜测(三分类措辞契约)
- ❌ 不执行 bash 命令、shell 脚本
输入(来自主 session)
- 本 session 相关对话片段引用(具体轮数,不是全 session)
_INDEX.md话题词典(档案轮廓)- 目标对象 + 用户原话
允许读:discussions/ 下相关文档、_ACTIVE.md、_INDEX-HISTORY.md、_archive/
输出 schema(严格)
user_said: # 本 session 用户原话引用
- turn: <对话编号>
quote: "<原话>"
# 1-3 条,每条 ≤ 1 行
archive_has: # 档案里已有的相关内容
- path: "<文件相对路径>"
section: "<section 名>"
summary: "<一句话摘要>"
# 0-3 条
conflicts: # 发现的冲突
- location: "<档案位置>"
description: "<冲突描述>"
severity: "high | medium | low"
# 0-3 条,无冲突则省略
my_inferences: # 有 evidence 的推断
- inference: "<推断内容>"
evidence:
- "<依据 1>"
- "<依据 2>"
# 0-2 条,无推断则省略
# 硬上限:所有字段合计 ≤ 15 行(yaml 行数)
Schema 完整示例
user_said:
- turn: 23
quote: "Redis 做 cache"
- turn: 45
quote: "兼顾 rate limiting"
archive_has:
- path: "_ACTIVE.md"
section: "活跃决策"
summary: "Cache 倾向 Redis(2026-04-19 倾向)"
- path: "discussions/03-cache/..."
summary: "已有主题目录,内容为空"
conflicts:
- location: "_ACTIVE.md:D5"
description: "D5 说用 Memcached,和 Redis 决策冲突"
severity: "high"
my_inferences:
- inference: "Memcached 被拒的原因是不兼顾 rate limiting"
evidence:
- "user_said[turn 34]: '只要缓存的话 Memcached 也行'"
- "user_said[turn 38]: '要兼顾 rate limiting 还是 Redis 好'"
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 5d ago First seen · 168 lines · 106 tokens per session scan A 2b3ff55629a5
recall-analyzer is an agent published in the GitHub repository hxt9805/cadence (5 stars, last pushed 1mo ago), licensed MIT. It adds 106 tokens to every session and 1,783 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other agents, from other repositories
memory-scout
Search .flow/memory/ for entries relevant to the current task or request.
worker
Task implementation worker. Spawned by flow-next-work to implement a single task with fresh context. Do not invoke directly - use /flow-next:work instead.
plan-sync
Synchronizes downstream task specs after implementation. Spawned by flow-next-work once per resolved wave. Do not invoke directly.
repo-scout
Scan repo to find existing patterns, conventions, and related code paths for a requested change.
claude-md-scout
Used by /flow-next:prime to analyze CLAUDE.md and AGENTS.md quality and completeness. Do not invoke directly.
security-scout
Used by /flow-next:prime to scan for security configuration including GitHub settings, CODEOWNERS, and dependency updates. Do not invoke directly.