simplify-and-harden

Agent Skill: Simplify & Harden

Install

gh skill install pskoett/pskoett-skills simplify-and-harden

For CI-only execution, use:

gh skill install pskoett/pskoett-skills simplify-and-harden-ci

Fallback using the Agent Skills CLI:

npx skills

Metadata

Rationale and Philosophy

When a coding agent completes a task, it holds peak contextual understanding of the problem, the solution, and the tradeoffs it made along the way. This context degrades immediately -- the next task wipes the slate. Simplify & Harden exploits that peak context window to perform two focused review passes before the agent moves on.

Most agents solve the ticket and stop. This skill turns "done" into "done well."

The operating philosophy is a deliberate "fresh eyes" self-review before moving on: carefully re-read all newly written code and all existing code modified in the task, and look hard for obvious bugs, errors, confusing logic, brittle assumptions, naming issues, and missed hardening opportunities. The goal is not to expand scope or rewrite the solution -- it is to use peak context to perform a disciplined first review pass while the agent still remembers the intent behind every change.

Best Use with Independent Review

This skill is a post-completion self-pass and does not replace an independent review pass.

Recommended flow:

1. Implement the task.
2. Run Simplify & Harden to clean, harden, and document non-obvious decisions.
3. Run an independent review pass for severity-ordered findings.
4. Merge only after both passes are addressed.

If the two disagree, treat the independent review findings as the external gate and either fix or explicitly waive findings.

Trigger Conditions

The skill activates automatically when ALL of the following are true:

• The agent has completed its primary coding task
• The agent signals task completion (exit code 0, PR ready, or equivalent)
• The diff contains a non-trivial code change (see definition below)
• The skill has not already run on this task (no re-entry loops)

Non-trivial code change definition

Treat a diff as non-trivial when either condition is true:

1. It touches at least one executable source file (for example: *.ts, *.tsx, *.js, *.jsx, *.py, *.go, *.rs, *.java, *.cs, *.rb, *.php, *.swift, *.kt, *.scala, *.sh) and includes either:
   • At least 10 changed non-comment, non-whitespace lines in executable source files, OR
   • At least one high-impact logic change (auth/authz checks, input validation, data access/query logic, external command execution, file path handling, network request handling, or concurrency control).
2. It changes configuration, workflow, permissions, deployment, or security policy in a way that can alter who may act, what executes, what is deployed, or which protections apply.

Treat the diff as non-trivial = false when it is docs-only, low-impact config-only, comments-only, formatting-only, generated artifacts only, or tests-only.

The skill does NOT activate when:

• The agent failed or was interrupted
• The change is documentation-only
• The change is tests-only
• The change is a generated file (lockfiles, build artifacts)
• The user explicitly skips it via --no-review or equivalent flag

Scope Constraints

Hard rule: Only touch files modified in this task.

The agent MUST NOT:

• Refactor adjacent code it did not modify
• Pursue "while I'm here" improvements outside the diff
• Introduce new dependencies or architectural changes
• Make speculative fixes based on patterns it noticed elsewhere

The agent SHOULD flag out-of-scope concerns in the summary output rather than acting on them.

Budget limits:

• Maximum additional changes: 20% of the original diff size (measured in lines changed)
• Maximum execution time: 60 seconds (configurable)
• If either limit is hit, the agent stops and outputs what it has with a budget_exceeded flag

Pass 1: Simplify

Objective: Reduce unnecessary complexity introduced during implementation.

Default posture: simplify, don't restructure. The primary goal of this pass is lightweight cleanup that is demonstrably behavior-preserving. Refactoring is the exception, not the rule.

Fresh-eyes start (mandatory): Before making any edits in this pass, re-read all code added or modified in this task with "fresh eyes" and actively look for obvious bugs, errors, confusing logic, brittle assumptions, naming issues, and missed hardening opportunities.

The agent reviews its own work and asks:

"Now that I understand the full solution, is there a simpler way to express this?"

Review Checklist

1. Dead code and scaffolding -- Did I leave behind debug logs, commented-out attempts, unused imports, or temporary variables from my iteration loop? Remove them.

2. Naming clarity -- Do function names, variables, and parameters make sense when read fresh? Names that made sense mid-implementation often read poorly after the fact. Rename them.

3. Control flow -- Can any nested conditionals be flattened? Can early returns replace deep nesting? Are there boolean expressions that could be simplified? Tighten them.

4. API surface -- Did I expose more than necessary? Could any public methods/functions be private? Reduce visibility.

5. Over-abstraction -- Did I create classes, interfaces, or wrapper functions that aren't justified by the current scope? Agents tend to over-engineer. Flag it, but don't restructure unless the win is significant.

6. Consolidation opportunities -- Did I spread logic across multiple functions or files when it could live in one place? Flag it, but only propose a refactor if the duplication is egregious and the consolidation is clean.

Simplify Actions

For each finding, the agent categorizes it as:

• Behavior-preserving fix (for example, removing a provably unused import or task-created dead code) -- applied automatically if within budget.
• Observable or uncertain change (including naming, control-flow, visibility, API, policy, permission, deployment, or security behavior) -- proposed for human approval even when the diff is small.
• Refactor (consolidation, restructuring, abstraction changes) -- proposed ONLY when the agent determines it is genuinely necessary or the benefit is substantial. A refactor is not the default action. The bar is: "Would a senior engineer look at this and say the current state is clearly wrong, not just imperfect?"

Refactor Stop Hook (mandatory):

Any change the agent classifies as a refactor triggers an interactive prompt. The agent MUST:

1. Describe what it wants to change and why
2. Show the before/after (or a clear description of the structural change)
3. Wait for explicit human approval before applying

The agent does not batch refactor proposals. Each refactor is presented individually so the human can approve, reject, or modify on a case-by-case basis.

If the human selects skip all refactors, the agent skips remaining refactor proposals and moves to the Harden pass. Skipped refactors still appear in the output summary as flagged with status skipped_by_user.

Only changes proven behavior-preserving do not trigger the stop hook. The category label never overrides observable behavior: a rename, branch rewrite, or visibility reduction may be externally significant and must be approved when its effect is uncertain.

Pass 2: Harden

Objective: Close security and resilience gaps while the agent still understands the code's intent.

The agent reviews its own work and asks:

"If someone malicious saw this code, what would they try?"

Review Checklist

1. Input validation -- Are all external inputs (user input, API params, file paths, environment variables) validated before use? Check for type coercion issues, missing bounds checks, and unconstrained string lengths.

2. Error handling -- Are catch blocks specific? Are errors logged with context but without leaking sensitive data? Are there any swallowed exceptions?

3. Injection vectors -- Check for SQL injection, XSS, command injection, path traversal, and template injection in any code that builds strings from external input.

4. Authentication and authorization -- Do new endpoints or functions enforce auth? Are permission checks present and correct? Is there any privilege escalation risk?

5. Secrets and credentials -- Are there hardcoded secrets, API keys, tokens, or passwords? Are connection strings parameterized? Check for credentials in log output.

6. Data exposure -- Does error output, logging, or API responses leak internal state, stack traces, database schemas, or PII?

7. Dependency risk -- Did the agent introduce new dependencies? If so, are they well-maintained, properly versioned, and free of known vulnerabilities?

8. Race conditions and state -- For concurrent code: are shared resources properly synchronized? Are there TOCTOU (time-of-check-to-time-of-use) vulnerabilities?

Harden Actions

For each finding, the agent categorizes it as:

• Patch (adding a validation check, escaping output, removing a hardcoded secret) -- applied automatically only when the approved acceptance contract already requires it and targeted verification can prove the result; otherwise request human approval
• Security refactor (restructuring auth flow, replacing a vulnerable pattern with a new approach, changing data handling architecture) -- ALWAYS requires human approval before proceeding

The same Refactor Stop Hook from the Simplify pass applies here. Security refactors are presented individually with the added context of severity and attack vector:

• Flagged as critical -- findings the agent cannot safely patch without human input (noted in output regardless of approval)
• Flagged as advisory -- hardening opportunities that are not active vulnerabilities

Security patches (not refactors) are prioritized over simplification changes when budget is constrained.

Pass 3: Document (Micro-pass)

Objective: Capture non-obvious decisions while the agent still remembers why it made them.

This is deliberately lightweight -- not a documentation pass, just decision capture.

Rules

• For any logic that requires more than 5 seconds of "why does this exist?" thought: add a single-line comment explaining the decision
• For any workaround or hack: add a comment with context and ideally a TODO with conditions for removal
• For any performance-sensitive choice: note why the current approach was chosen over the obvious alternative
• Maximum: 5 comments added per task. This is not a documentation sprint.

Pass 4: Re-verify

Re-verification is mandatory before signaling completion.

1. If Simplify & Harden changed any file, re-run the smallest existing tests, checks, linters, or behavioral acceptance that cover those edits.
2. If it made no edits, the immediately preceding verify-gate evidence may be reused when it still covers the reviewed diff.
3. If verification fails, return to implementation with the diagnostics; do not emit a successful completion.
4. If no applicable verification exists, record not-run, set review_followup_required: true, and do not claim the task is fully verified.

Output Schema

The skill produces a structured summary appended to the task output:

Set review_followup_required to true when any unresolved finding remains (critical/advisory flags, skipped or timed-out refactor proposals), when budget_exceeded is true, or when verification is not pass. Set it to false only when no follow-up is required.

Self-Improvement Integration (Learning Loop)

Simplify & Harden feeds its recurring quality/security findings into the self-improvement skill so repeated issues can become durable prompt rules.

After each run:

1. Normalize each finding into a pattern_key:
   • Simplify examples: simplify.dead_code, simplify.naming, simplify.control_flow
   • Harden examples: harden.input_validation, harden.authorization, harden.error_handling
2. Emit those pattern candidates in simplify_and_harden.learning_loop.candidates.
3. Hand off candidates to self-improvement, which logs or updates entries in .learnings/LEARNINGS.md (instead of creating duplicate one-off notes).
4. Mark candidates as promotion-ready when they cross the recurrence threshold: >= 3 occurrences across >= 2 distinct tasks in a 30-day window.
5. Promote promotion-ready patterns into the agent context/system prompt files (CLAUDE.md, AGENTS.md, .github/copilot-instructions.md, or equivalent) to reduce repeat issues.

This keeps Simplify & Harden focused on per-task cleanup/hardening while self-improvement owns cross-task memory and promotion.

Execution Model

This skill is for general coding-agent sessions where a human can approve refactors in-line.

Behavior:

• Refactor proposals are shown one at a time with clear rationale
• The agent pauses and waits for [approve], [reject], [show diff], or [skip all refactors]
• Only provably behavior-preserving fixes and acceptance-backed security patches are applied automatically

For CI pipelines and headless automation, use simplify-and-harden-ci.

Agent Context File References

To activate this skill, reference it in your agent context file.

Agent-specific copy-paste snippets are in references/agent-context-snippets.md. Load only the snippet for your active agent to keep context lean.

Core invariants for any agent integration:

1. Scope lock -- only files modified in the current task
2. Budget cap -- 20% max additional diff
3. Simplify-first posture -- cleanup is the default, refactoring is the exception
4. Refactor stop hook -- structural changes always require human approval
5. Four phases -- simplify, harden, document, re-verify (in that order)
6. Post-edit verification -- rerun applicable checks after any review edit
7. Structured output -- summary of applied, approved, rejected, flagged, and verified items

Precaution: some agents may not reliably pause for approval in high-autonomy modes. Validate this behavior before production use.

Agent Compatibility

This skill is designed to work with any coding agent that follows a task-based workflow. It is not tied to any specific agent framework or product.

Programmatic integration (agents with skill/hook APIs):

• Claude Code, GitHub Copilot Workspace, Codex, Opencode, OpenClaw, Cursor Agent, Windsurf, Aider, SWE-Agent, OpenHands, Devin, and any agent exposing a task completion lifecycle event

Prompt-based integration (chat-based agents without formal skill APIs):

• Any LLM-based coding assistant that accepts post-task instructions -- the skill's logic can be injected as a follow-up prompt after the agent signals completion

The output schema is agent-agnostic YAML. Consuming tools only need to parse the structured summary.

Integration Notes

This skill is agent-agnostic. It hooks into any coding agent that exposes a task completion lifecycle event. The examples below are generic -- adapt them to your agent's specific API.

Agent Integration

The skill hooks into the agent's task completion lifecycle. Suggested integration pattern:

Agents that support this skill should implement the following interface:

• Access to the diff produced by the completed task
• A list of modified files with full content
• The ability to present interactive prompts (for interactive mode)
• An output channel for the structured summary (stdout, PR comment, or equivalent)

Prompt-based Integration

For agents that don't support programmatic skill hooks (e.g., chat-based coding agents like Claude Code, Cursor, Copilot Chat), this skill can be implemented as a post-task prompt injection:

CI Pipeline Variant

For GitHub Actions or other CI/headless usage, run simplify-and-harden-ci.

Configuration

Design Decisions

Why post-completion and not continuous? Continuous review during implementation creates feedback loops that slow the agent down and can cause oscillation (simplify, then re-complicate, then re-simplify). Post-completion gives the agent a stable codebase to review against.

Why simplify-first, not refactor-first? Agents love to refactor. Given permission to "improve" code, they will restructure it. But most post-task improvements are cosmetic: a dead import, a bad name, a needlessly deep conditional. These account for 80%+ of the value with near-zero risk. Refactoring carries real risk -- it can introduce bugs, break tests, and bloat diffs. By making simplification the default and refactoring the exception, the skill delivers consistent value without surprise rewrites. The bar for a refactor should be "this is genuinely wrong" not "this could be slightly better."

Why a budget? Without constraints, agents will use review passes as license for unbounded refactoring. The 20% rule keeps the skill focused: improve what you built, don't rebuild it.

Why separate simplify from harden? They require different mindsets. Simplify asks "is this the clearest expression of my intent?" while Harden asks "how could this be exploited?" Conflating them leads to mediocre results on both. Running them sequentially also lets us prioritize security fixes when budget is tight.

Why the document micro-pass? Agents are terrible at documenting their reasoning unprompted. Humans reviewing agent-generated code consistently report that the biggest friction is understanding why a choice was made. Five comments is a trivial cost for enormous review-time savings.

Future Considerations

• Team calibration: Allow teams to weight the review checklist (e.g., "we care more about injection vectors than naming")
• Diff-aware context loading: For large codebases, intelligently load only the files and symbols relevant to the diff rather than the full project
• Cross-skill composition: Simplify & Harden could feed into a "PR Description" skill that uses its summary to auto-generate meaningful PR descriptions