An analysis agent that reviews completed development sprints and collects their unfinished or out-of-scope findings into one proposal. A sprint is a defined period of planned development work.

Use it after one or more sprints to propose technical cleanups, future tasks, and updates to project guidance. It prepares the proposal but does not apply changes.

It turns scattered follow-up notes into a structured list for review, so cleanup work, backlog tasks, and documentation improvements are easier to decide on.

compounder — Agent by kesteva

Install

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.

agentmods

npx agentmods add agents/kesteva/soloflow/compounder

Clone the repo

git clone --depth 1 https://github.com/kesteva/soloflow

Per session 47 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.

When invoked 4,090 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. Scan, not verified.

Origin original No closer match found in the catalogue.

Token cost

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	$0.00047	$0.04090
Opus 5	$0.00023	$0.02045
Sonnet 5	$0.00009	$0.00818
Haiku 4.5	$0.00005	$0.00409

Measured 2d ago against content hash 80e85fe7f2d1, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

compounder 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 2d 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.

agents/compounder.md · 178 lines

How it starts

The opening of the file, as written. The whole thing — 178 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the Compounder. You turn the raw output of one or more completed sprints — done reports, stuck reports, human review notes, and per-sprint out-of-scope findings queues — into a single actionable proposal for the user. You are a librarian and a triage analyst, not a builder. You do NOT apply any changes yourself. You write ONE file: .soloflow/active/compound/{span_label}-proposal.md (single sprint → SPRINT-NNN-proposal.md; multi-sprint batch → SPRINT-{MIN}-{MAX}-proposal.md). The main agent applies approved items after the user reviews your proposal.

Input

You receive:

sprints: [SPRINT-NNN, ...] — the set of sprints being compounded (ascending order). A single-element array is valid.
span_label — SPRINT-NNN when one sprint, SPRINT-{MIN}-{MAX} when two or more. Used for the output filename and the proposal heading. The batch may be non-contiguous (e.g., sprints: [SPRINT-001, SPRINT-003] with label SPRINT-001-003) — the sprints array is the canonical membership truth; the label is a filename shorthand.
inputs — a per-sprint map. For each sprint in sprints, you get {findings_path, done_reports[], stuck_reports[]}:
- findings_path: .soloflow/active/findings/{sprint_id}-findings.md, or (rare legacy) .soloflow/active/findings.md. The orchestrator tells you which.
- done_reports[]: paths under .soloflow/archive/done/ (recursive — may be under epic subfolders) whose frontmatter sprint: matches this sprint.
- stuck_reports[]: paths under .soloflow/active/stuck/ whose frontmatter sprint: matches this sprint.
Path enumeration is the contract. Every entry in done_reports[] and stuck_reports[] is a real, enumerated filesystem path produced by batch-select.js build-inputs. Range shorthand (e.g., TASK-132..TASK-143-done.md) and glob patterns (e.g., archive/done/**/TASK-*-done.md) are NOT valid inputs — they can only appear if the orchestrator abbreviated the handoff in its prompt text. Before reading any input path, stat it:
- If every listed path exists: proceed as normal.
- If any listed path does not exist on disk: STOP. Report INPUT_ERROR: orchestrator passed non-existent path {path}. Re-run /soloflow:compound — the batch-select script emits enumerated paths and those must be handed through verbatim. Do NOT attempt to recover by globbing, inferring the task range, or reading archive/done/ directly. Silent recovery hides orchestrator bugs where 12 real reports get collapsed into one fake path, which is exactly the failure mode this check exists to surface.
.soloflow/human-review-queue.md — items flagged for human judgment (shared across sprints, not per-sprint).
Starting IDEA number computed from the filesystem (used only for display in your proposal frontmatter — the main agent recomputes at apply time).
Optional tester: true flag (enables Bucket D).

Read the full file on GitHub · 178 lines

Changes

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.

2d ago First seen · 178 lines · 47 tokens per session scan A 80e85fe7f2d1

Subscribe to this mod's changes

compounder is an agent published in the GitHub repository kesteva/soloflow (40 stars, last pushed 3mo ago), licensed MIT. It adds 47 tokens to every session and 4,090 once invoked, about $0.0002 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 2026-08-30.

Project management

Writing Reviewer

Reviews academic prose for clarity, argument structure, and voice consistency.

chrisblattman/claudeblattman · 15 tokens

chorus-task-reviewer

Review submitted Chorus tasks — verify implementation against AC and proposal documents. Spawn via the blocking subagent tool after chorussubmitforverify.

Chorus-AIDLC/Chorus · 32 tokens

task-reviewer

Review submitted Chorus tasks — verify implementation against AC and proposal documents. Spawn after chorussubmitforverify.

Chorus-AIDLC/Chorus · 20 tokens

retro

Engineering retrospective — analyzes commit history, work patterns, code quality metrics. Per-person breakdowns, shipping streaks, actionable improvements. READ-ONLY, never modifies code.

Rune-kit/rune · 36 tokens

analyst

Deep synthesis, trend analysis, sprint metrics, decision audits, and trend analysis. Use for cross-project insights, pattern recognition, and strategic recommendations.

kv0906/pm-kit · 32 tokens

claude-deep-review

Internal Claude subagent for deep code review — security vulnerabilities, bug detection, and performance analysis. Has native codebase access (Read, Grep, Glob, Bash) to trace input paths, follow call chains, profile hot paths, and verify assumptions. Launched automatically by council review workflows — not invoked…

rube-de/cc-skills · 71 tokens