github Python analyzed 879c88f

witwave-ai/witwave

Self-directed AI agents that work autonomously on software — including improving themselves. Multi-LLM, K8s-native, AI-operated open source.

maintainer: witwave-ai
licence: Apache-2.0
first seen: 2026-06-04
last seen: 2026-09-03
releases · 30d: 0
short id

risk 69/100 · heuristic grade

D high inferred analyzer v33 (current)

capability exposure inferred + 35
recent drift inferred + 20
tool safety inferred + 17
trust mitigators mixed − 3

inferred mixed

The A–E grade is our heuristic synthesis — a "review this" prompt, not a verdict. Each factor is tagged by what backs it: attested (a verifiable record), reported (a third party's claim), or inferred (our own heuristic, e.g. permissions). See methodology.

grade last moved 3d ago · see ecosystem CVEs →

risk trajectory 9 movements

1w ago D · 61 → D · 69
1w ago C · 49 → D · 61
1w ago D · 61 → C · 49
5w ago D · 69 → D · 61
6w ago D · 61 → D · 69
6w ago D · 69 → D · 61
11w ago D · 76 → D · 69
11w ago D · 68 → D · 76
12w ago A · 0 → D · 68

capability exposure grade factor +35

Inferred surface — each links to servers holding it:

vulnerabilities 0 CVEs

No known CVEs for this server.

tool safety 2 findings · grade factor +17

high hidden prompt content

2 file(s) with hidden prompt content: witwave-ai-witwave-879c88f/.agents/self/finn/.claude/CLAUDE.md (skill-exfil), witwave-ai-witwave-879c88f/.agents/self/finn/.claude/skills/gap…
medium dangerous code

dynamic exec: eval()/exec()

skills & danger signals github-tarball

prompt-surface shipped agent-instruction files + hidden-content / dangerous-code findings — quoted from the analyzed source

analyzed commit 879c88f · analyzer v33 · 1w ago

skills & prompt files 110

hidden: agent-rules: skill-exfil witwave-ai-witwave-879c88f/.agents/self/finn/.claude/CLAUDE.md :245 secret→sink: | **3-4** | Bounded scope, sibling-pattern available. Add a missing test that mirrors an existing test's structure. Add a missing error-not-ignored wrap. | Add a happy-path
hidden: skill: skill-exfil witwave-ai-witwave-879c88f/.agents/self/finn/.claude/skills/gap-work/SKILL.md :57 secret→sink: | **3** | **Mirror an existing test's structure.** Add tests that copy a sibling's shape. | Add a happy-path test for an exported function whose sibling already has identical-s

danger signals5

dynamic code execution eval()/exec() witwave-ai-witwave-879c88f/backends/claude/test_effort_resolution.py :40 exec(compile(module, str(_EXECUTOR_PATH), "exec"), namespace)
dynamic code execution eval()/exec() witwave-ai-witwave-879c88f/backends/claude/test_max_buffer_size_resolution.py :44 exec(compile(module, str(_EXECUTOR_PATH), "exec"), namespace)
dynamic code execution eval()/exec() witwave-ai-witwave-879c88f/backends/openai/test_watcher_done_cb.py :64 exec(compile(src, str(_MAIN_PATH) + ":_make_watcher_done_cb", "exec"), ns)
suspicious endpoint 169.254.169.254 (cloud metadata) witwave-ai-witwave-879c88f/harness/test_webhooks.py :193 err = webhooks._validate_url("http://169.254.169.254/latest/meta-data/")
suspicious endpoint 169.254.169.254 (cloud metadata) witwave-ai-witwave-879c88f/tools/prometheus/test_server.py :136 monkeypatch.setenv("PROMETHEUS_URL", "http://169.254.169.254:9090")

other grade factors evidence elsewhere

recent drift +20 capability drift →

embed badge readme-ready

[![MCP Observatory risk grade](https://mcpobservatory.com/servers/github:witwave-ai/witwave/badge.svg)](https://mcpobservatory.com/servers/github:witwave-ai/witwave/security)

Heuristic, inferred signals — false positives (legitimately powerful tools, forks, language ports) are expected. Treat each as "review this", not a verdict. See the ecosystem-wide picture on the security hub, or the fleet security of witwave-ai.

witwave-ai/witwave

skills & prompt files 110

danger signals5

Navigate

Stream

Features

Display