05-guidelines

A report auditing whether a coding workflow followed its documented rules and user feedback across multiple development cycles.

In plain words
What is it for?
Use it to review process compliance, missing tests, merge behavior, and recurring quality problems in an agent-driven development workflow.
Why use it?
It reveals repeated rule violations, gaps between written instructions and actual behavior, and defects that reached the main codebase.

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/proggarapsody/bitbottle/05-guidelines
Clone the repo
git clone --depth 1 https://github.com/proggarapsody/bitbottle
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 4,683 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 $0.00000 $0.04683
Opus 5 $0.00000 $0.02341
Sonnet 5 $0.00000 $0.00937
Haiku 4.5 $0.00000 $0.00468

Measured 2d ago against content hash 4c2b061cddde, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

05-guidelines 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.

auto-iter/reports/analysis-2026-06-02-cycles-158-187/agents/05-guidelines.md · 204 lines

How it starts

The opening of the file, as written. The whole thing — 204 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Guidelines Audit — Cycles 153–187

Dimension: Adherence to documented workflow rules and explicit user feedback
Generated: 2026-06-02
Analyst: review-agent (Claude Sonnet 4.6)


TL;DR

Across 30 cycles (153–187), six distinct compliance dimensions were audited. Three rules showed sustained violations: metric emission collapsed to 0–2 steps/cycle for cycles 168–187 (vs. 10 expected), cycles 168–169 had confirmed dispatch violations, and the integration-test requirement (ARCHITECTURE.md tier-2) was missing from TDD subagent prompts, causing DJ BLOCKERs in cycles 185 and 187. The auto-merge "race" (5/10 cycles 178–187) is not a rule violation: the old documented rule in quickref.md explicitly accepted the "arm on CI green, fix via follow-up PR" tradeoff; the new "arm after DJ" rule was written reactively to memory but never landed in tracked files, creating a documentation-vs-behavior gap. Cycle 179 (BACKLOG-MIGRATION standalone chore) is a genuine but mitigated violation: the §4 gate in pre-merge-mechanical.sh only catches single-file commits and silently passed the two-file doc-only PR.


Compliance Matrix

Rule Source Cycles Checked Verdict Violations
1. Mandatory step compliance (lock → preflight → PRD → TDD → DJ → pre-merge → merge → release → cleanup) quickref.md §Metrics schema; README.md §0–§10 153–187 FAIL Metric emission collapsed: 0 steps logged for cycles 169–177; 2 steps for 178–187 (vs. ~10 expected). Cycles 162, 163, 165 ran TDD work but were never appended to cycles.jsonl.
2. BACKLOG→SHIPPED discipline (move in same feat commit, not standalone chore) README.md §4; quickref.md anti-patterns; pre-merge-mechanical.sh §4 153–187 CONDITIONAL FAIL Cycle 179 (BACKLOG-MIGRATION, PR #625): standalone chore(backlog) PR touching only docs/backlog/BACKLOG.md and docs/backlog/SHIPPED.md, with no feat code. Gate §4 did not catch it (§4 only blocks file_count == 1; cycle 179 had file_count == 2). Earlier: cycle 147 had confirmed standalone chore PR #507 (standalone BACKLOG flip noted in stream 145–154 report).
3. Auto-merge gating ("arm after DJ returns SHIP, not on CI green") feedback_auto_merge_race.md (memory, 2026-06-01) 176–187 NOT APPLICABLE at time of cycles Old quickref.md rule (last updated 2026-05-29) explicitly endorsed arm-on-CI-green with follow-up PR. Cycles 176–177 (2 follow-up PRs) and 178–187 (5 follow-up PRs) followed the documented behavior. The new rule exists only in memory and has not been landed in quickref.md, README.md, or auto-iter.md. This is a documentation gap, not a behavioral failure.
4. Integration-test requirement (*_integration_test.go required per new command) feedback_integration_test_required.md (memory, 2026-06-01); docs/ARCHITECTURE.md §Test tiers 185, 187 FAIL Cycle 185 (REPO-SYNC, PR #643): missing integration test — DJ BLOCKER, fix PR required. Cycle 187 (COMMIT-SEARCH, PR #651): missing search_integration_test.go — DJ BLOCKER, fix PR #7bfe6b6 required. The requirement exists in ARCHITECTURE.md but is absent from the TDD subagent prompt in .claude/commands/auto-iter.md.
5. Orchestrator-dispatch rule (all code writes go to subagents, orchestrator is shell-only) feedback_orchestrator_inline_work.md; quickref.md anti-patterns 168–187 PASS with 2 violations corrected Cycle 168 (HOST-INFO): orchestrator applied DJ dead-branch fix inline. Cycle 169 (API-PARITY): full TDD cycle done inline, no DJ run. Both logged as dispatch violations in stream 168–177 report. Corrected by cycle 172; cycles 172–187 show no dispatch violations. Cycles 178–187: dispatch violations = 0.
6. Release-halt-in-stream rule (skip ship? halt in stream mode) feedback_release_halt_stream.md 168–187 PASS Both streams (168–177 and 178–187): zero release halts recorded. Stream 168–177: user pre-authorized no halts; stream 178–187: same. No step2_halt or step2_halt2 metric lines appear in cycles.jsonl for either stream. Correct.
7. No internal hostnames in public artifacts feedback_no_internal_hostnames.md (2026-05-19) 153–187 PRE-EXISTING DEBT pkg/cmd/mcp/handlers_test.go contains git.moscow.alfaintra.net at lines 166, 168, 190. However, this was introduced on 2026-04-26 (before the rule was established 2026-05-19) in feat: add pipeline and branch commands. No cycle in 153–187 added new hostname instances; subsequent modifications (cycles 174–175) inherited the existing lines. Not a new violation from the audited window — existing technical debt.
8. Agent-rules-location rule (new algorithm rules in tracked auto-iter/ or docs/workflows/, not memory) feedback_agent_rules_location.md Post-178–187 FAIL Two new rules generated as feedback exist only in gitignored memory: (a) feedback_auto_merge_race.md — the "arm after DJ" gate rule; (b) feedback_integration_test_required.md — the explicit integration test requirement. Neither has been landed in quickref.md, README.md, or auto-iter.md. Per the rule: "any rule that should govern multiple agent harnesses must live under tracked auto-iter/ or docs/workflows/."

Read the full file on GitHub · 204 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. 2d ago First seen · 204 lines · 0 tokens per session scan A 4c2b061cddde

Subscribe to this mod's changes

05-guidelines is an agent published in the GitHub repository proggarapsody/bitbottle (3 stars, last pushed 15d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 4,683 tokens. 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.