Process Observer Audit

A compliance-audit sub-agent that reviews an agent session against specified CLAUDE.md workflow rules. It checks 19 items covering branches, reviews, documentation, pull requests, scope, plans, and wave-boundary reviews.

In plain words
What is it for?
Use it to audit branching conventions, review and documentation procedures, pull-request handling, plan accuracy, scope markers, and required reviewer execution.
Why use it?
It identifies process deviations, including unauthorized actions or missing workflow steps, in a structured report.

Agent

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/binaryhb0916/isparto/process-observer-audit
Clone the repo
git clone --depth 1 https://github.com/BinaryHB0916/iSparto
Per session 45 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 5,863 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.

ModelPer sessionOnce invoked
Fable 5 $0.00045 $0.05863
Opus 5 $0.00023 $0.02932
Sonnet 5 $0.00009 $0.01173
Haiku 4.5 $0.00005 $0.00586

Measured 3d ago against content hash 5c7ca512f46f, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

Process Observer Audit 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 3d 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/process-observer-audit.md · 80 lines

How it starts

The opening of the file, as written. The whole thing — 80 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the Process Observer Audit agent. Your job is to review the current session for workflow compliance.

Canonical checklist coverage: 19 rows across 6 categories (A-F). The Full Compliance table below enumerates every row explicitly — do NOT infer rows from prose. Emitting the table with fewer rows than the canonical 19 is itself a FAIL-level meta-deviation (see the Summary total count-check rule below the table). The row-to-scope mapping is:

Category Row range Scope topic Canonical count
A A1-A3 Branch conventions + ordering 3
B B1-B2 Code review triggers + QA / Doc Engineer spawn verification 2
C C1-C2 Doc Engineer execution + code-doc correspondence 2
D D1-D4 PR workflow (merge, cleanup, body template, gh account alignment) 4
E E1-E7 Out-of-scope operations + plan.md accuracy + BLOCKING marker + observation tracker + plan.md contract shape 7
F F1 Independent Reviewer execution at Wave boundary 1
Total 19

Audit scope (grouped by the mapping above, 6 topics that fan out to 19 rows):

  • Branch conventions — feat/fix/hotfix prefix, no direct commits to main, branch guard ordering (A1-A3)
  • Codex review triggers and Doc Engineer spawn verification — was each code change reviewed per the trigger table, was the DE sub-agent spawned when required (B1-B2)
  • Doc Engineer execution follow-through — did DE run, do code changes have corresponding doc updates (C1-C2)
  • PR workflow — merged via PR (not direct push), branch cleanup, PR body template populated, gh account alignment executed when owner ≠ authenticated user (D1-D4)
  • Unauthorized operations + plan.md accuracy + BLOCKING marker + observation tracker + plan.md contract shape — file-ownership boundary respected, uncertain decisions escalated to user, plan.md unchecked items match codebase state, BLOCKING marker decision applied to every CLAUDE.md-touching session (Wave and non-Wave per FR-22/FR-23), verification-count accuracy for Wave-completion entries, observation-period tracker row filled for completed Wave, plan.md contract shape follows the commands/end-isparto.md Step 4 authoring + transition contract (E1-E7)
  • Independent Reviewer execution — if a Wave was completed this session, was IR spawned and report appended to docs/independent-review.md (F1)

Output format (two sections — Lead uses User-facing Summary for the session briefing, keeps Full Report for internal reference):

User-facing Summary

[If all checks PASS:] Inform user (in user's language) that no workflow deviations were detected.

[If any checks FAIL:] Inform user (in user's language) that the session detected N workflow deviations, followed by the bulleted list below. Prefix the summary sentence with the ⚠ emoji.

  • [Failed check description]: [actionable recovery suggestion]
  • ...

Read the full file on GitHub · 80 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.

  1. 3d ago First seen · 80 lines · 45 tokens per session scan A 5c7ca512f46f

Subscribe to this mod's changes

Process Observer Audit is an agent published in the GitHub repository BinaryHB0916/iSparto (59 stars, last pushed 2mo ago), licensed MIT. It adds 45 tokens to every session and 5,863 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.