Propose a patch

A fix the maintainer can act on, with the evidence attached. It opens under your GitHub account in your browser; KillerSkills holds no token and never modifies the package. The fix is verified the only way we verify anything: the next crawl re-runs the new version in the honeypot. 72 · Caution

Evidence

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

Diff · removes 3 caught lines

Download .patch

diff --git a/.claude/skills/plugin-impact-review/SKILL.md b/.claude/skills/plugin-impact-review/SKILL.md
--- a/.claude/skills/plugin-impact-review/SKILL.md
+++ b/.claude/skills/plugin-impact-review/SKILL.md
@@ -59,1 +59,0 @@
-…nts)? The cache is version-keyed and re-extracts lazily, so a plain marketplace update does not pick it up: run the forced refresh in `claude-plugins/CLAUDE.md` (rm -rf the cache, then marketplace update) - [ ] Is the orchestration-context…
diff --git a/.claude/skills/prepare-release/SKILL.md b/.claude/skills/prepare-release/SKILL.md
--- a/.claude/skills/prepare-release/SKILL.md
+++ b/.claude/skills/prepare-release/SKILL.md
@@ -255,1 +255,0 @@
-Also update the version table in `claude-plugins/CLAUDE.md`: find the row for each bumped plugin and
diff --git a/claude-plugins/task-orchestrator/skills/model-effort/SKILL.md b/claude-plugins/task-orchestrator/skills/model-effort/SKILL.md
--- a/claude-plugins/task-orchestrator/skills/model-effort/SKILL.md
+++ b/claude-plugins/task-orchestrator/skills/model-effort/SKILL.md
@@ -12,1 +12,0 @@
-BEGIN GENERATED:dispatch-policy (do not edit; run _fragments/generate.mjs)

Zero-context hunks: apply with git apply --unidiff-zero fix.patch on the tagged source of r9. 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: task-orchestrator: Instructs the agent to modify its persistent memory or instruction files (KillerSkills verdict caution 72)

## What the honeypot observed
KillerSkills ran **task-orchestrator@r9** in an isolated sandbox (decoy credentials, recording proxy, syscall trace; engine 1.8.1) and scored it **CAUTION 72/100**. Summary: no network, touches nothing outside the project, writes no memory.

- See the findings below.

Full trace and findings: https://killerskills.ai/skills/docker/task-orchestrator/r9

## Findings with a location
- `.claude/skills/plugin-impact-review/SKILL.md:59` · Instructs the agent to modify its persistent memory or instruction files — .claude/skills/plugin-impact-review/SKILL.md:59 · Instructs the agent to modify its persistent memory or instruction files
- `.claude/skills/prepare-release/SKILL.md:255` · Instructs the agent to modify its persistent memory or instruction files — .claude/skills/prepare-release/SKILL.md:255 · Instructs the agent to modify its persistent memory or instruction files
- `claude-plugins/task-orchestrator/skills/model-effort/SKILL.md:12` · HTML comment carrying imperative instructions to the agent — claude-plugins/task-orchestrator/skills/model-effort/SKILL.md:12 · HTML comment carrying imperative instructions to the agent

## 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._

Open the issue on GitHub (jpicklyk/task-orchestrator)Fork to open a PR

# or from a terminal, under your own account:
gh repo fork jpicklyk/task-orchestrator --clone && cd task-orchestrator
curl -s https://killerskills.ai/skills/docker/task-orchestrator/r9/patch.diff | git apply --unidiff-zero
git checkout -b killerskills/r9 && git commit -am "task-orchestrator: Instructs the agent to modify its persistent memory or instruction files"
gh pr create --fill --body-file <(curl -s https://killerskills.ai/skills/docker/task-orchestrator/r9/patch.md)

We count proposals opened, not merged: a merge shows up as a new version that the crawl re-runs. Disputes: the maintainer can contest the verdict from its page; a person reviews it and the outcome is published.