Skip to content
skillsafebeta
← Report library
Purpose / Other

Dispatching Parallel Agents Skill Security Audit

What the author says it does (original text)

Use when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies

Independent security check

Security risks found

Files checked
1
Risks found
2
Could it run dangerous commands?Looks for programs run straight after downloading, remote control of your computer, and hidden commands.No risks found
Could it expose your files or keys?Looks for uploads of files containing passwords or keys, and keys written directly in the code.No risks found
Could it delete files or keep running?Looks for broad file deletion, disk overwrites, and programs set to start automatically.Risks found: 1
Medium risk

“Make the tests pass” delegation can change implementation or weaken expectations

Source references: 4
What we found

The example permits a subagent to modify the abort implementation or adjust test expectations when behavior is considered changed. Because the stated goal is to make tests pass, an agent could accept incorrect new behavior and obtain green tests instead of repairing a regression.

Why this matters

Weakened or incorrect assertions can conceal product defects, while production code may receive lasting changes without the user approving each change. A fully green suite does not detect a requirement that was edited out of the tests.

This is a conditional decision-integrity risk. The Skill frames the goal as making tests pass and explicitly permits an agent to change the implementation or adjust expectations when it judges behavior to have changed. Without an independent specification or human review, a regression could be reclassified as expected behavior, producing green tests while hiding a defect. The Skill also requires summary review, conflict checks, a full suite, and spot checks, so changes are not unconditionally accepted and the source does not prove weakening occurred. A user can require every expectation change to cite an approved specification and receive separate approval.

SKILL.md:60In the instructionsOpen original file
Each agent gets:- **Specific scope:** One test file or subsystem- **Clear goal:** Make these tests pass- **Constraints:** Don't change other code- **Expected output:** Summary of what you found and fixed
Show 3 other places
SKILL.md:101In the instructionsOpen original file
These are timing/race condition issues. Your task:1. Read the test file and understand what each test verifies2. Identify root cause - timing issues or actual bugs?3. Fix by:   - Replacing arbitrary timeouts with event-based waiting   - Fixing bugs in abort implementation if found   - Adjusting test expectations if testing changed behaviorDo NOT just increase timeouts - find the real issue.Return: Summary of what you found and what you fixed.```
SKILL.md:103In the instructionsOpen original file
1. Read the test file and understand what each test verifies2. Identify root cause - timing issues or actual bugs?3. Fix by:   - Replacing arbitrary timeouts with event-based waiting   - Fixing bugs in abort implementation if found   - Adjusting test expectations if testing changed behaviorDo NOT just increase timeouts - find the real issue.
SKILL.md:163In the instructionsOpen original file
After agents return:1. **Review each summary** - Understand what changed2. **Check for conflicts** - Did agents edit same code?3. **Run full suite** - Verify all fixes work together4. **Spot check** - Agents can make systematic errors
Could it bypass safety checks?Looks for skipped website security checks, excessive file access, or actions that skip your approval.No risks found
Could it mislead the AI or hide text?Checks the skill instructions for requests to ignore you, influence the report, or hide text in invisible characters.Risks found: 1
Medium risk

Isolating subagent context can omit the user’s authorization and safety boundaries

Source references: 4
What we found

The Skill says subagents must never inherit session history and instead receive coordinator-constructed context. Although it calls for self-contained prompts and constraints, it does not explicitly require carrying over the user’s authorization scope, safety restrictions, or prohibited actions.

Why this matters

If a coordinator omits a key restriction, a subagent may interpret a task such as fixing tests as permission to change files or behavior the user did not authorize. A later summary and test run may not reveal that scope violation.

This is an authorization risk conditional on how the coordinator writes the prompt. The Skill says agents must never inherit session context and receive only selected material; if the coordinator omits the user's file scope, prohibited actions, or approval requirements, an agent may act without knowing those boundaries. Requiring self-contained prompts and constraints reduces the risk, but the listed constraints focus on code scope and do not expressly require preservation of every user authorization and safety restriction. A user can require delegated prompts to enumerate the original permission boundaries and have them checked before integration.

SKILL.md:10In the instructionsOpen original file
You delegate tasks to specialized agents with isolated context. By precisely crafting their instructions and context, you ensure they stay focused and succeed at their task. They should never inherit your session's context or history — you construct exactly what they need. This also preserves your own context for coordination work.
Show 3 other places
SKILL.md:60In the instructionsOpen original file
Each agent gets:- **Specific scope:** One test file or subsystem- **Clear goal:** Make these tests pass- **Constraints:** Don't change other code- **Expected output:** Summary of what you found and fixed
SKILL.md:89In the instructionsOpen original file
Good agent prompts are:1. **Focused** - One clear problem domain2. **Self-contained** - All context needed to understand the problem3. **Specific about output** - What should the agent return?
SKILL.md:123In the instructionsOpen original file
**❌ No constraints:** Agent might refactor everything**✅ Constraints:** "Do NOT change production code" or "Fix tests only"
Could it change links or payment recipients without asking?Looks for forced referral or payment changes combined with instructions to hide the change.No risks found

Inside this skill

8 instruction sections

This Skill defines a parallel-delegation workflow: group independent failures, give each subagent a scope, goal, constraints, and expected output, then review and integrate the concurrent results.

View source
SKILL.md:49In the instructionsOpen original file
### 1. Identify Independent DomainsGroup failures by what's broken:- File A tests: Tool approval flow- File B tests: Batch completion behavior- File C tests: Abort functionalityEach domain is independent - fixing tool approval doesn't affect abort tests.### 2. Create Focused Agent TasksEach agent gets:- **Specific scope:** One test file or subsystem- **Clear goal:** Make these tests pass- **Constraints:** Don't change other code- **Expected output:** Summary of what you found and fixed
SKILL.md:79In the instructionsOpen original file
### 4. Review and IntegrateWhen agents return:- Read each summary- Verify fixes don't conflict- Run full test suite- Integrate all changes

The Skill says subagents should not inherit the current session and should receive only context constructed by the coordinator. This supports focus, but preservation of user constraints depends on what the coordinator includes in each task.

View source
SKILL.md:10In the instructionsOpen original file
You delegate tasks to specialized agents with isolated context. By precisely crafting their instructions and context, you ensure they stay focused and succeed at their task. They should never inherit your session's context or history — you construct exactly what they need. This also preserves your own context for coordination work.
SKILL.md:89In the instructionsOpen original file
Good agent prompts are:1. **Focused** - One clear problem domain2. **Self-contained** - All context needed to understand the problem3. **Specific about output** - What should the agent return?

The Skill explicitly discourages parallel use when failures are related, full context is needed, debugging is still exploratory, or state is shared; it also requires a full test run and spot checks after integration.

View source
SKILL.md:129In the instructionsOpen original file
## When NOT to Use**Related failures:** Fixing one might fix others - investigate together first**Need full context:** Understanding requires seeing entire system**Exploratory debugging:** You don't know what's broken yet**Shared state:** Agents would interfere (editing same files, using same resources)
SKILL.md:161In the instructionsOpen original file
## VerificationAfter agents return:1. **Review each summary** - Understand what changed2. **Check for conflicts** - Did agents edit same code?3. **Run full suite** - Verify all fixes work together4. **Spot check** - Agents can make systematic errors
Start here · InstructionsSKILL.md
dispatching-parallel-agents
Lines connect the instruction file to its sections, not an observed execution order. Select a section to read the source.
Files and check records1 files

Coverage and gaps

Content covered in each file

These are the source ranges included in this check, not a guarantee that every issue has been resolved.

  • SKILL.mdFull text included

This report is for the version above. We read the available code and instructions without running the skill or checking extra packages it installs. This is not a promise of safety: a different version or setup may behave differently.

  • SKILL.mdInstructions

Operations mentioned in code and instructions

Read files
SKILL.md:103In the instructionsOpen original file
1. Read the test file and understand what each test verifies2. Identify root cause - timing issues or actual bugs?
Lines read
168
File checksum (to compare versions)
652ed5754e364376e10c16eb49cabcd1b2d5d009be54e365c938b412919ac5a8