bounds-retro
Examine a bounded set of accessible engineering work for repeated corrections and wasted effort, then propose evidence-backed improvements to structure…
Skill by boundsj
Chalk Labs
Need help deploying bounds-retro?
Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/boundsj-bounds-retro
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/bounds-retro
What's inside
- Skill name
bounds-retro- Description
- Examine a bounded set of accessible engineering work for repeated corrections and wasted effort, then propose evidence-backed improvements to structure, checks, tooling, or navigation.
- License
- MIT
- Instructions
- 35 lines in SKILL.md · view on GitHub
Files (4)
- LICENSE.txt
- NOTICE.md
- SKILL.md
- sources.md · references/
SKILL.md (first lines)
# Bounds retro Improve the environment the next run works in, so the user corrects less. This is not a code review or a memory system. ## Set the boundary Use the work the user names: a session, T3 thread, delegated review, PR, or branch. Default to the current conversation, diff, and check results. For a named past session, read its own record; [sources](references/sources.md) covers host logs. State the boundary and missing evidence. Do not search unrelated histories. If compaction removed the middle of the current session, say so and suggest a retro pointed at its log. ## Find the cost Human attention is the primary signal: user messages that correct, re-ask, repeat a constraint, or unblock; reviewer findings the implementer should have caught; early stops that left a closable gap. Then look for slow navigation, expensive or repeated tool calls, retries, misleading checks, and missing information. When work spanned providers, models, or delegated reviewers, attribute each moment to its agent. The fix may belong in a review brief or host setting rather than the repository. Every candidate must trace to a specific turn, command, or finding; discard untraceable advice. Separate repeated patterns from single incidents. A smooth session may have nothing worth changing. ## Choose where the fix lands Inspect existing check commands, hooks, and CI first. Then prefer the cheapest durable mechanism: 1. Structure that removes the mistake: one owner, clear interfaces, fewer competing paths. 2. A deterministic check for a mechanical violation. Repair or wire an existing check before adding one. No guardrail at all is itself a finding. 3. Reviewer-read standards for a judgement call. Review has the least context pressure; keep standards out of always-loaded instructions. 4. A navigation pointer from an already-read file for slow discovery; wider access, such as a teed log, for missing information; a streamlined tool for expensive calls. 5. A short standing instruction only for what nothing above can carry. Keep `AGENTS.md`/`CLAUDE.md` mostly to pointers; flag no-ops and misplaced steering for removal. ## Report Rank by cost to the user's attention, not loudness. For each: moment → cause → change and where it lands → how to validate. Keep each a short, phone-readable block; return few unless asked. Mark deletions. Say what one retro cannot judge, such as whether an older rule blocks good work. Invent no time savings. ## Act Propose by default. Implement only when already authorized, within the identified project. Prove a new check fails on a representative mistake and accepts valid work before it blocks merges. Provider and host changes are proposals for the user. Do not edit global instructions, add hooks, install skills, publish histories, or create tracker tasks merely because they might help.
From boundsj/bounds-skills, MIT.
More from boundsj/bounds-skills
- 0.570score
bounds-architect
boundsj
Design state ownership, types, interfaces, and module shape before implementing a nontrivial change, comparing a few meaningful alternatives with…
SkillAFreshness: AClaude Code +9 - 0.570score
bounds-automate-me
boundsj
Create or update a scoped skill that captures how the user likes agents to work, from their explicit preferences and bounded evidence of repeated decisions.
SkillAFreshness: AClaude Code +9 - 0.570score
bounds-mode
boundsj
Coordinate a task with Bounds workflows when the user explicitly asks for bounds-mode.
SkillAFreshness: AClaude Code +9 - 0.567score
bounds-why
boundsj
Investigate why existing code, a design, a threshold, or a past decision has its current shape, separating cited evidence from inference, alternatives, and…
SkillAFreshness: AClaude Code +9 - 0.567score
bounds-create-verification
boundsj
Create and execute a project-local verification recipe and small feature map using the project's existing tools to launch, health-check, drive, capture…
SkillAFreshness: AClaude Code +9 - 0.567score
bounds-recall
boundsj
Recover prior goals, decisions, constraints, and unfinished work from accessible agent sessions and project records, then verify current state.
SkillAFreshness: AClaude Code +9 - 0.567score
bounds-unslop
boundsj
Edit prose the user asks to tighten so it reads direct and concrete, preserving meaning, commitments, numbers, uncertainty, citations, quotes, and code.
SkillAFreshness: AClaude Code +9 - 0.561score
bounds-review
boundsj
Review a pinned diff or change for correctness, requirement compliance, and maintainability, separating actionable findings from preferences and unverified…
SkillAFreshness: AClaude Code +9 - 0.561score
bounds-ketchup
boundsj
Brief the user on what changed in the current conversation since their last substantive message: current state, evidence, blockers, decisions they own, and…
SkillAFreshness: AClaude Code +9 - 0.561score
bounds-plan
boundsj
Turn an engineering idea or change request into a concise, executable plan with scope, retained behavior, acceptance checks, and real dependencies.
SkillAFreshness: AClaude Code +9 - 0.561score
bounds-blast-radius
boundsj
Trace what an exact code or configuration change could break beyond its diff, such as callers, shared state, client contracts, caches, serialization, and…
SkillAFreshness: AClaude Code +9 - 0.561score
bounds-debug
boundsj
Reproduce an observed bug or performance regression, isolate its cause, fix it within scope, and verify the user's failing behavior with a repeatable check.
SkillAFreshness: AClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Oct 6, 2026 (25 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/boundsj/bounds-skills
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from boundsj
- 0.570score
bounds-mode
boundsj
Coordinate a task with Bounds workflows when the user explicitly asks for bounds-mode.
SkillAFreshness: AClaude Code +9 - 0.570score
bounds-automate-me
boundsj
Create or update a scoped skill that captures how the user likes agents to work, from their explicit preferences and bounded evidence of repeated decisions.
SkillAFreshness: AClaude Code +9 - 0.570score
bounds-architect
boundsj
Design state ownership, types, interfaces, and module shape before implementing a nontrivial change, comparing a few meaningful alternatives with…
SkillAFreshness: AClaude Code +9 - 0.567score
bounds-create-verification
boundsj
Create and execute a project-local verification recipe and small feature map using the project's existing tools to launch, health-check, drive, capture…
SkillAFreshness: AClaude Code +9 - 0.567score
bounds-recall
boundsj
Recover prior goals, decisions, constraints, and unfinished work from accessible agent sessions and project records, then verify current state.
SkillAFreshness: AClaude Code +9 - 0.567score
bounds-unslop
boundsj
Edit prose the user asks to tighten so it reads direct and concrete, preserving meaning, commitments, numbers, uncertainty, citations, quotes, and code.
SkillAFreshness: AClaude Code +9
Done for you
Rather not set it up yourself?
Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.
See how we build AI agents →Read the case studies
Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.
- launches shipped
- 50+
- launches shipped
- media placements
- 500+
- media placements
- founders on every account
- 2
- founders on every account
Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block
Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.
Get bounds-retro running in your business
Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.
Cite this page
bounds-retro skill. Agent Mart. https://agentmart.dev/skills/boundsj-bounds-retro (last updated 2026-10-24)