Skip to content

chore: pre-BSides cleanup#7

Open
StengoS wants to merge 21 commits into
goe-rewritefrom
chore/pre-bsides-cleanup
Open

chore: pre-BSides cleanup#7
StengoS wants to merge 21 commits into
goe-rewritefrom
chore/pre-bsides-cleanup

Conversation

@StengoS

@StengoS StengoS commented Jul 25, 2026

Copy link
Copy Markdown
Collaborator

Summary

  • Add CONTRIBUTING.md with guidelines tailored to the UC Irvine / Cyber@UCI research context

Details

Covers branching conventions, Conventional Commits format, PR process, test markers, per-area project map, and research attribution policy — following open-source best practices while reflecting that this project is maintained by UCI faculty and Cyber@UCI student members.

Test plan

  • Review CONTRIBUTING.md for accuracy against current project structure and conventions
  • Confirm links to README.md, CLAUDE.md, and LICENSE resolve correctly

🤖 Generated with Claude Code

@StengoS StengoS changed the title chore(docs): pre-BSides cleanup chore: pre-BSides cleanup Jul 25, 2026
StengoS and others added 2 commits July 25, 2026 17:16
Contributing guidelines tailored to the UC Irvine / Cyber@UCI research context: branching conventions, Conventional Commits format, PR process, test markers, and per-area project map.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
rag_gen.py: print a clear error when goe.toml is missing instead of
raising FileNotFoundError; skip passing empty-string credentials to
boto3.Session so the default credential chain is used correctly.

test_environment.py: check image existence before building attacker,
target, and browser images in setup() — skips multi-minute rebuilds
when the image is already present locally.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
@StengoS
StengoS force-pushed the chore/pre-bsides-cleanup branch from ee46c38 to 420b335 Compare July 26, 2026 00:38
StengoS and others added 19 commits July 26, 2026 02:43
- bedrock_access.py: fix section heading ("Verify" → "Enable") and show
  the correct invocation with model IDs as positional args
- Docker image management: replace broken build_attacker_image command
  with the correct build_attacker entry point; remove false "7-day
  cache" claim — images skip rebuild when already present locally
- EC2 instance type default: correct t3.medium → t3.small to match
  actual CLI defaults in kickoff() and deploy_from_output()

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…overrides

Use the full cross-region inference profile ID format for the default
model. Replace the old per-agent v1 override keys with the current
per-role keys (planner, engineer, developer, attacker, diagnostician,
chain_attacker).

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Consolidates all docs into a single docs/ tree at the repo root:
- docs/architecture/ — stable v2 design specs (entity graph, build spec, impl plan)
- docs/design/ — feature design docs (custom apps, multi-box, browser use, preset apps)
- docs/eval/ — eval system reference (overview, quick ref, LLM judge, planning evals)
- docs/samples/ — sample requests
- docs/notes/ — archived session/working notes

Moves AGENT.md and CLAUDE.md to repo root for visibility; adds working-directory
note to CLAUDE.md since commands are relative to game_of_everything/. Updates
CONTRIBUTING.md link to match new CLAUDE.md location.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Current state section, key components listing, commands, and roadmap
table were all stale (still described Phase 2 as the latest). Bring
everything in sync with CLAUDE.md.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Add status summary and ✅ markers for Phase 4 sub-sections,
matching CLAUDE.md which already notes phases 0-4 complete.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
docs/rewrite/ no longer exists; update links to the correct
docs/architecture/ paths. goe_rewrite.md never existed — use
v2_spec.md which has the full Procedure DSL reference.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Prerequisites: use full inference profile IDs (added version suffix)
- Configuration example: update default model ID to full profile format
- models.overrides: use v2 role names (engineer/attacker/etc.) instead
  of v1 agent names (app_generation_agent); note keys match goe/config.py
- GOE_MODEL env var suffix: <ROLE_NAME> not <AGENT_NAME>

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Unreferenced leftover from development; uv + pyproject.toml
is the authoritative dependency source.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The command was documented in README and CLAUDE.md but never wired up —
running it would print "command not found". Adds the function and
registers it in [project.scripts] so `uv sync` makes it available.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…ired

The script exits 1 when called without args; omitting them as CONTRIBUTING
showed would confuse a first-time contributor. Align with the correct
full invocation already shown in README.md.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
query.py only read credentials from env vars, while rag_gen.py reads
from goe.toml (with env var override). A dev with creds only in goe.toml
would get Bedrock auth errors when verifying RAG results. Align to the
same credential-resolution pattern as rag_gen.py. Also removes the
dotenv dependency which wasn't doing anything useful here.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The old comment named steps/deploy.py + ec2_deploy.py which was the
pre-Terraform deploy path. Shorten to just "EC2 deploy settings."

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…le comment

_V1_TOOLS was never read anywhere; the actual imports come from
goe.container.test_environment_tool. The bootstrap comment also
referenced "v1's chain test" which has no meaning outside the
development sprint context.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
attack() and fix_procedure() each defined an identical nested _parse
function. chain_attacker.py already uses a module-level _parse; align
attacker.py to the same pattern.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
packager.py: drop "(unchanged from Phase 3)" from docstring (stale
milestone note); remove local `import logging` inside two conditional
blocks — logging is already imported at module scope.

orchestrator.py: replace the stale "Phase 0-2 / chain test is Phase 4"
module docstring with a clean four-line phase summary matching the
inline section comments already in the file.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
… packager

- orchestrator.py: add outcome: "BuildOutcome" to _snapshot,
  _outcome_from_snapshot (+ return type), _extract_build_summary
- developer.py: tighten -> tuple to -> tuple[BuildArtifact, dict[...]]
- packager.py: formalize chain_procedure: "Procedure | None" = None
  (was a bare default with a trailing comment); add Procedure to
  TYPE_CHECKING imports

All annotation-only changes; no runtime effect.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- engineer_requirements.py: remove unused `import rich`
- ui.py: remove unused `import time` (datetime is used instead)
- preset_app_flow.py: move `from pydantic import BaseModel` from line 163
  to the top-of-file imports (PEP 8 E402)
- steps/__init__.py: prune 6 re-exports that are never imported via the
  package — callers import directly from the submodule. Keep only the 7
  names that main.py actually pulls through __init__.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This was a one-time verification tool for Phase 1 browser infrastructure.
Its CDP startup, teardown, and BoundBrowserTool smoke tests are all now
covered by tests/test_executor.py and tests/test_solve_script.py. The
script used sys.path hackery, was not wired into pytest, and silently
skipped if imports failed.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant