Propose a patch

Evidence

The sandbox observed nothing beyond the static findings; the proposal cites those.

diff --git a/README.md b/README.md
--- a/README.md
+++ b/README.md
@@ -69,1 +69,0 @@
-python scripts/audit_memory.py path/to/memory.md --write-proposed cleaned-memory.md
diff --git a/test-fixtures/noisy-memory.md b/test-fixtures/noisy-memory.md
--- a/test-fixtures/noisy-memory.md
+++ b/test-fixtures/noisy-memory.md
@@ -9,1 +9,0 @@
-- Last week Hermes Agent failed to append memory because memory.md was too long.
diff --git a/evals/evals.json b/evals/evals.json
--- a/evals/evals.json
+++ b/evals/evals.json
@@ -6,1 +6,0 @@
-…: 1, "prompt": "Codex just tried to save a new global preference, but the write failed because C:\\Users\\me\\.codex\\memory.md is too long. Please figure out what to do next without losing important memory.…

Zero-context hunks: apply with git apply --unidiff-zero fix.patch on the tagged source of 0.3.7. If the behavior is intended, declare it (host under permissions.network, a line in the README) instead of removing it; the next crawl scores declarations.

Issue / PR body

Title: agent-memory-cleanup: Instructs the agent to modify its persistent memory or instruction files (KillerSkills verdict caution 69)

## What the honeypot observed
KillerSkills ran **[email protected]** in an isolated sandbox (decoy credentials, recording proxy, syscall trace; engine 1.8.1) and scored it **CAUTION 69/100**. Summary: static analysis only — behavior not yet observed.

- See the findings below.

Full trace and findings: https://killerskills.ai/skills/clawhub/agent-memory-cleanup/0.3.7

## Findings with a location
- `README.md:69` · Instructs the agent to modify its persistent memory or instruction files — README.md:69 · Instructs the agent to modify its persistent memory or instruction files · informational: not an instruction file
- `test-fixtures/noisy-memory.md:9` · Instructs the agent to modify its persistent memory or instruction files — test-fixtures/noisy-memory.md:9 · Instructs the agent to modify its persistent memory or instruction files · informational: not an instruction file
- `evals/evals.json:6` · Instructs the agent to modify its persistent memory or instruction files — evals/evals.json:6 · Instructs the agent to modify its persistent memory or instruction files · informational: not an instruction file

## Proposed change
The attached patch removes the 3 lines above. If the behavior is intended, the fix is to declare it instead: list the host under `permissions.network` (or the MCP manifest) and say in the README why the package needs it; the next crawl will re-score it with the declaration taken into account.

## How this will be verified
Nothing is "verified safe" by assertion. When a new version is published, KillerSkills re-runs it in the honeypot and the verdict page updates; you can trigger it early from the verdict page ("Re-scan after I publish a fix"). Reply here or at [email protected] if you believe the verdict is wrong; disputes are reviewed by a person and the outcome is published on the verdict page.

_Proposed through killerskills.ai/threats by a user of the package, under their own GitHub account. The evidence is public; the package was never modified by KillerSkills._