Skip to content

Profiles: sourced access facts for 15 local tools (Q4 plan E3) - #348

Merged
YugantM merged 3 commits into
mainfrom
e3-access-facts
Oct 9, 2026
Merged

YugantM merged 3 commits into
mainfrom
e3-access-facts

Conversation

@YugantM

@YugantM YugantM commented Oct 6, 2026 •

Copy link
Copy Markdown
Owner

Q4 plan E3. Line-reviewed by the owner on 9 Oct; every edit applied as proposed. The diff only touches safety_facts.json.

What it adds

55 facts for the 15 local tools behind most of the remaining safety searches in the E1 audit (#344). Each fact has a source and a label: PUBLISHER (the project's own words), CHECKED (observed by us on 9 Oct) or IN CODE (read from a pinned release of its source):

Profile Highlights
Ollama API binds 127.0.0.1:11434 by default; the official Docker image binds 0.0.0.0 (IN CODE); cloud prompts are processed but not stored or trained on (its words); OSV: 16 advisories (1 critical, 11 high), 3 in 12 months
AnythingLLM telemetry on unless turned off; 25 maintainer advisories, 23 in 12 months
OpenCode every tool, edits and shell included, runs without asking by default; asks outside the project and before reading .env (IN CODE; its docs wrongly say .env is denied)
Desktop Commander its own SECURITY.md: reads/writes files and runs any command; guardrails "not a security sandbox"; tool arguments logged unredacted locally; 2 low OSV advisories
Open WebUI Tools/Functions run arbitrary Python on the server (admins only, per its docs); 165 maintainer advisories, 162 in 12 months
Cherry Studio data local; prompts go straight to your provider, except via its built-in model service relay
Blender MCP any Python runs by default (safe mode opt-in); socket server has no auth; telemetry may train AI models; 1 OSV advisory
Qwen Code Auto mode default; usage statistics on by default (IN CODE), though its notice calls them optional
Headroom, OpenShell, Odysseus, OnGrid, CKAN MCP, Aider, Agent Reach access model, telemetry defaults, advisories, security policy

Review rules applied (from #347)

  • A PUBLISHER fact says only what its source says. Defaults are read from code (pinned tag and line) when the docs are silent or wrong.
  • Advisory lines carry severities. GitHub counts are paginated; OSV counts are de-duplicated across aliases (GHSA, GO-, PYSEC-). Products with no GitHub advisories were checked on OSV, which added Aider and Blender MCP.
  • Agent Reach's README is in Chinese. The facts are close translations, and the first one says so.
  • lastmod_diff: 7 profiles re-dated locally. The other 8 are pending_signals in stale local data but scored in prod, so all 15 will show and re-date there. No other pages move.

./scripts/gates.sh green; safety_fact_problems returns none.

🤖 Generated with Claude Code

@YugantM
YugantM deleted the branch main October 8, 2026 10:29
@YugantM YugantM closed this Oct 8, 2026
@YugantM YugantM reopened this Oct 8, 2026
@YugantM
YugantM changed the base branch from e2-safety-facts to main October 8, 2026 10:29
yugant and others added 3 commits October 9, 2026 15:51
Adds 52 facts for the local tools behind most of the remaining "is X
safe?" searches (#344): Ollama, AnythingLLM, OpenCode, Desktop Commander,
Open WebUI, Cherry Studio, Blender MCP, Headroom, Agent Reach, OpenShell,
CKAN MCP, Qwen Code, Odysseus, OnGrid and Aider. What each can reach
(shell, files, network binding, plugins that run code, approval
defaults), where data goes (telemetry defaults and opt-outs, local
storage, model providers) and security practice (published advisories
counted from GitHub or OSV, SECURITY.md, private reporting).

Data only, on top of the E2 mechanism. lastmod_diff vs E2: 7 pages
locally; the other 8 are pending in local data but scored in prod.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Severities on every maintainer advisory count (re-counted with pagination
on 8 Oct; Cherry Studio's 4 are all critical), and no "no advisories
published" lines, which can read as "no vulnerabilities". OSV-based counts
for Ollama (31) and Desktop Commander (2) re-checked, unchanged.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Re-checked all 52 facts on 9 Oct against their sources, the paginated
GitHub advisory counts, OSV, and the code wherever a fact states a default.

- OpenCode: the docs say .env reads are denied by default, but v1.18.35
  asks; the default is now an IN CODE fact, plus a publisher fact on rules.
- Ollama: OSV's 31 entries are 16 advisories (GHSA and Go ids alias each
  other); severities added. New IN CODE fact: the Docker image binds
  0.0.0.0.
- OSV advisories added for Aider (2, low) and Blender MCP (1, medium).
- Qwen Code: usage statistics are on by default in code (IN CODE); the
  provider line lists all four sign-in methods and drops the inference.
- Cherry Studio's built-in model service relays prompts; Desktop
  Commander logs arguments unredacted; Blender MCP's socket has no auth,
  its safe mode, and telemetry may train AI models; Aider commits skip
  pre-commit hooks; Headroom's off switch is HEADROOM_OFFLINE=1;
  AnythingLLM's switch is in the sidebar.
- checked dates move to 2026-10-09; 55 facts in all.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@YugantM
YugantM marked this pull request as ready for review October 9, 2026 13:55
@YugantM
YugantM merged commit ab6528b into main Oct 9, 2026
3 checks passed
@YugantM
YugantM deleted the e3-access-facts branch October 9, 2026 14:00
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