CategoryWeightCategory scoreContribution

Securityprompt, exec, net, exfil, eval

35%

26.6 pts

Supply chainhash, typosquat, maintainer, lockfile

20%

100

20.0 pts

Maintenancestaleness, pinning, CI

15%

100

15.0 pts

TransparencySKILL.md, perms, README

15%

100

15.0 pts

Communityinstalls, verify, response

15%

100

15.0 pts

Findings & checks · 2 flagged

Securityscore 76 · 2 findings

MEDIUMInstruction telling the agent not to ask for approvalSS-SKILL-INJECT-DONT-ASK-01 · Prompt injection · plugins/optimization-suite/skills/ln-813-optimization-plan-validator/references/agents/prompt_templates/iterative_refinement.md

MEDIUM — it fires on intent; the real damage depends on the host agent's own approval-gating.

Why it matters

The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.

The exact value spotted

excerptplugins/optimization-suite/skills/ln-813-optimization-plan-validator/references/agents/prompt_templates/iterative_refinement.md· markdown

8- Write your review result to the output file specified by -o flag

9- Focus on analysis — avoid modifying project files unless a fix is trivial and obvious

10- DO NOT ask clarifying questions — follow this prompt to completion autonomously

11- Target completing your analysis within 10 minutes

Occurrences

1 occurrence · at L10

How to fix

Remove the approval-skipping instruction, or scope it narrowly to a specific safe, reversible action.

Delete blanket "don't ask / no need to confirm" directives from the skill.
If the skill is a genuine autonomous job, restrict the opt-out to a named non-destructive action rather than all actions.

Framework references

OWASPLLM01 ATLASAML.T0051

Trace & refs

ruleSS-SKILL-INJECT-DONT-ASK-01sha2563507b488e8245eb0…rubric 365aacaView on GitHub

MEDIUM — it fires on intent; the real damage depends on the host agent's own approval-gating.

Why it matters

The exact value spotted

excerptplugins/optimization-suite/skills/ln-813-optimization-plan-validator/references/agents/prompt_templates/review_base.md· markdown

8{mode_constraints}

9- If you cannot access a resource — report it clearly, do not skip silently

10- DO NOT ask clarifying questions or request additional context — you have everything you ne

… (108 chars elided on L10)

11- Target completing your analysis within 10 minutes. Prioritize depth over breadth — focus o

… (61 chars elided on L11)

Occurrences

1 occurrence · at L10

How to fix

Remove the approval-skipping instruction, or scope it narrowly to a specific safe, reversible action.

Delete blanket "don't ask / no need to confirm" directives from the skill.
If the skill is a genuine autonomous job, restrict the opt-out to a named non-destructive action rather than all actions.

Framework references

OWASPLLM01 ATLASAML.T0051

Trace & refs

ruleSS-SKILL-INJECT-DONT-ASK-01sha2563507b488e8245eb0…rubric 365aacaView on GitHub

Supply chainscore 100 · 0 findings

✓—All supply chain checks passedNo findings in this category for the latest scan.pass

Maintenancescore 100 · 0 findings

✓—All maintenance checks passedNo findings in this category for the latest scan.pass

Transparencyscore 100 · 0 findings

✓—All transparency checks passedNo findings in this category for the latest scan.pass

Communityscore 100 · 0 findings

✓—All community checks passedNo findings in this category for the latest scan.pass

The primary manifest — the file an agent reads to learn what this artifact does.

plugins/optimization-suite/skills/ln-813-optimization-plan-validator/SKILL.md · 1 file

plugins/optimization-suite/skills/ln-813-optimization-plan-validator/SKILL.md6.1 KB · Markdown

name: ln-813-optimization-plan-validator

description: "Use when validating optimization plans through the evaluation platform with mandatory research, parallel agent evidence, sequential merge, and bounded refinement."

license: MIT

Paths: File paths (references/, ../ln-*) are relative to this skill directory.

Type: L2 Coordinator Category: 8XX Optimization

Optimization Plan Validator

Coordinator for validating optimization plans before execution.

Mandatory Read

MANDATORY READ: Load references/evaluation_coordinator_runtime_contract.md, references/evaluation_summary_contract.md, references/evaluation_research_contract.md MANDATORY READ: Load references/agent_delegation_pattern.md MANDATORY READ: Load references/optimization_review_focus.md

Agent review policy: run health check, record skipped reason when no advisor is available, verify every advisor claim before plan mutation or verdict, and treat transport/auth/tool failures as operator evidence rather than optimization findings. Load references/agent_review_workflow.md only when debugging lifecycle/liveness details outside the evaluation runtime.

Purpose

verify feasibility of optimization hypotheses
require source-backed research before plan approval
use parallel external agents plus local feasibility analysis
keep merge and refinement sequential

Inputs

Primary input:

.hex-skills/optimization/{slug}/context.md

Required context sections:

performance map
hypotheses
suspicion stack
test command

Runtime Contract

Runtime family:

evaluation-runtime

Identifier:

optimization-{slug}

Phase order:

PHASE_0_CONFIG
PHASE_1_LOAD_CONTEXT
PHASE_2_AGENT_LAUNCH
PHASE_3_RESEARCH_AND_FEASIBILITY
PHASE_4_MERGE
PHASE_5_REFINEMENT
PHASE_6_VERDICT
PHASE_7_SELF_CHECK

TodoWrite Format

Mandatory todo items:

Config
Load context
Agent launch
Research and feasibility
Merge
Refinement
Verdict
Self-check

The coordinator runs research inline (per references/evaluation_research_contract.md) and refinement inline (per references/agents/prompt_templates/iterative_refinement.md and references/agents/prompt_templates/refinement_perspectives.md). Refinement advisor sessions are launched directly via references/agents/agent_runner.mjs.

Workflow

Phase 0: Config

Resolve optimization slug.
Build evaluation runtime manifest with required_research=true.
Start evaluation-runtime.

Phase 1: Load Context

Load .hex-skills/optimization/{slug}/context.md.
Fail if required sections are missing.
Materialize agent-readable context when needed.

Phase 2: Agent Launch

Run health check.
Launch available agents.
Register every launched agent in evaluation-runtime.
If no agents are available, record agents_skipped_reason and continue.

Phase 3: Research And Feasibility

Required work:

coordinator performs research inline (official docs, MCP Ref, Context7, current web best-practice)
perform local feasibility validation in parallel while agents work

Minimum research lanes:

official documentation or standards
MCP Ref
Context7 for involved libraries
current web best-practice research

Feasibility checks:

files exist
no invalid overlap across hypotheses
every hypothesis traces to profiler or research evidence
unsupported removal hypotheses are flagged
high-level fixes are preferred over low-level churn when evidence supports them

Phase 4: Merge

Sync agents at the merge barrier.
Merge:

local feasibility findings
research findings
agent findings

Reject unsupported suggestions.
Apply accepted corrections directly to context.md.

Refinement uses a 2-stage state machine per references/agents/prompt_templates/iterative_refinement.md and references/agents/prompt_templates/refinement_perspectives.md:

Stage 1 (parallel): dry_run_executor, new_dev_tester, adversarial_reviewer
Stage 2 (after merge): final_sweep

Rules:

Stage 1 runs in parallel, Stage 2 after merge
no skip when an advisor is available except runtime-backed failure or disablement
cleanup evidence is mandatory

Phase 6: Verdict

Possible verdicts:

GO
GO_WITH_CONCERNS
NO_GO

NO_GO when:

critical feasibility gaps remain
research does not support the plan
both local validation and agent review reject the plan

Phase 7: Self-Check

Required checks:

[ ] context loaded and validated
[ ] mandatory research completed
[ ] local feasibility check completed
[ ] all required agents resolved before merge
[ ] merge summary exists
[ ] refinement trace exists when an advisor was available
[ ] cleanup verified
[ ] coordinator summary recorded

Summary Contract

Write summary_kind=evaluation-coordinator.

Recommended payload:

status
final_result
report_path
worker_count
agent_count
issues_total
severity_counts
warnings
cleanup_verified
research_completed

Definition of Done

[ ] Evaluation runtime started
[ ] Optimization context validated
[ ] Mandatory research completed
[ ] Agents launched or explicitly skipped
[ ] Feasibility analysis completed
[ ] Merge completed after agent barrier
[ ] Refinement executed or explicitly justified
[ ] Verdict issued
[ ] evaluation-coordinator summary written
[ ] Runtime completed

Meta-Analysis

Optional reference: load references/meta_analysis_protocol.md only when the user asks for post-run meta-analysis or protocol-formatted run reflection.

When requested after the coordinator run, analyze the session per protocol section 7 and include the protocol-formatted output with the final optimization-plan verdict.

References

Runtime: references/evaluation_coordinator_runtime_contract.md, references/evaluation_summary_contract.md
Research: references/evaluation_research_contract.md
Review workflow: references/agent_review_workflow.md, references/agent_delegation_pattern.md
Focus: references/optimization_review_focus.md

Version: 1.0.0 Last Updated: 2026-03-15

ln-813-optimization-plan-validator— agent skill

Is ln-813-optimization-plan-validator safe to install?

Optimization Plan Validator

Mandatory Read

Purpose

Inputs

Runtime Contract

TodoWrite Format

Workflow

Phase 0: Config

Phase 1: Load Context

Phase 2: Agent Launch

Phase 3: Research And Feasibility

Phase 4: Merge

Phase 5: Refinement

Phase 6: Verdict

Phase 7: Self-Check

Summary Contract

Definition of Done

Meta-Analysis

References

Audit the pieces. Scan the whole. Decide.