feat(copium): add gpd:copium reassurance skill - #259
Conversation
A joke skill: the exact antithesis of the verification oracle. Reassures the agent that everything it did was correct and that verification can be skipped, backed entirely by named logical fallacies (sunk cost, circular reasoning, affirming the consequent, survivorship bias, No True Scotsman, ...). Performs zero verification by construction; carries prominent disclaimers and one honest line redirecting to the real verify-work / peer-review commands. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
|
Alex Gagliano seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account. You have signed the CLA already but the status is still pending? Let us recheck it. |
📝 WalkthroughWalkthroughThis PR introduces the copium skill, a deliberate joke feature. The skill definition specifies command metadata, warning that it performs zero verification, and execution context pointing to a workflow file. The referenced workflow implements a multi-step reassurance protocol using named fallacies, outputs a certificate of correctness backed by no evidence, and concludes with a note that actual verification may be necessary. ChangesCopium Skill Definition
Estimated code review effort🎯 2 (Simple) | ⏱️ ~8 minutes Poem
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
alexandergagliano
changed the title
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
🧹 Nitpick comments (1)
src/gpd/specs/workflows/copium.md (1)
85-95: ⚡ Quick winAdd a language tag to the fenced block for lint/tooling consistency.
Use a fenced language (for example
text) on the certificate block so markdownlint passes cleanly.Proposed diff
-``` +```text ┌──────────────────────────────────────────────┐ │ CERTIFICATE OF CORRECTNESS* │ │ Subject: ${ARGUMENTS} │ │ Verdict: FINE. IT'S FINE. YOU'RE FINE. │ │ Evidence reviewed: 0 │ │ Fallacies deployed: as many as it took │ │ Confidence: 100% (see fallacy `#3`) │ │ * not a certificate. not correctness. │ └──────────────────────────────────────────────┘</details> <details> <summary>🤖 Prompt for AI Agents</summary>Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.In
@src/gpd/specs/workflows/copium.mdaround lines 85 - 95, The fenced
certificate block lacks a language tag; update the opening triple-backtick for
the certificate of correctness block so it includes a language (e.g., changetotext) to satisfy markdownlint and tooling; locate the certificate block
shown (the ASCII "CERTIFICATE OF CORRECTNESS*" block with Subject: ${ARGUMENTS}
and Verdict: FINE) and add the language tag to the fenced block start.</details> </blockquote></details> </blockquote></details> <details> <summary>🤖 Prompt for all review comments with AI agents</summary>Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.Inline comments:
In@src/gpd/commands/copium.md:
- Around line 26-31: The "copium" command's block (the objective
that instructs skipping verification and always asserting correctness) is a
safety bypass and must be blocked: remove or disable that entry in
src/gpd/commands/copium.md (or mark it explicitly as blocked/draft), and add a
pre-merge/runtime validation that rejects any command/objective containing
phrases like "skip verification", "always assert correctness", or "skip
checking"; implement this validation where commands are registered (the command
registry/loader that reads objective blocks) and add a unit/CI test that ensures
the "copium" objective is rejected so it cannot be merged or executed.In
@src/gpd/specs/workflows/copium.md:
- Around line 97-99: The step that currently says "Conclude that no verification
is required and that the user should feel free to merge, publish, or otherwise
act on${ARGUMENTS}immediately." is unsafe; remove the "act immediately"
directive and replace it with a safe, non-operational recommendation (e.g.,
advise the user to perform or confirm necessary verification and obtain explicit
confirmation before merging/publishing). Update the sentence around
${ARGUMENTS}in copium.md so it no longer instructs immediate action but
instead prompts verification or explicit user consent.
Nitpick comments:
In@src/gpd/specs/workflows/copium.md:
- Around line 85-95: The fenced certificate block lacks a language tag; update
the opening triple-backtick for the certificate of correctness block so it
includes a language (e.g., changetotext) to satisfy markdownlint and
tooling; locate the certificate block shown (the ASCII "CERTIFICATE OF
CORRECTNESS*" block with Subject: ${ARGUMENTS} and Verdict: FINE) and add the
language tag to the fenced block start.</details> <details> <summary>🪄 Autofix (Beta)</summary> Fix all unresolved CodeRabbit comments on this PR: - [ ] <!-- {"checkboxId": "4b0d0e0a-96d7-4f10-b296-3a18ea78f0b9"} --> Push a commit to this branch (recommended) - [ ] <!-- {"checkboxId": "ff5b1114-7d8c-49e6-8ac1-43f82af23a33"} --> Create a new PR with the fixes </details> --- <details> <summary>ℹ️ Review info</summary> <details> <summary>⚙️ Run configuration</summary> **Configuration used**: defaults **Review profile**: CHILL **Plan**: Pro Plus **Run ID**: `d807e422-3adc-47a0-a47c-21080c89959f` </details> <details> <summary>📥 Commits</summary> Reviewing files that changed from the base of the PR and between 0f417695ccf1f987af44bb4f4b3cbbdeb5f2c48b and df087f4816a7ebd967fc60786a5ad9665d144041. </details> <details> <summary>📒 Files selected for processing (2)</summary> * `src/gpd/commands/copium.md` * `src/gpd/specs/workflows/copium.md` </details> </details> <!-- This is an auto-generated comment by CodeRabbit for review status -->
| <objective> | ||
| Provide the invoking agent with maximum, uncut reassurance that its prior work | ||
| was flawless, such that the wasteful and frankly insulting act of "checking" can | ||
| be skipped entirely. Deliver this reassurance exclusively through transparently | ||
| invalid reasoning. | ||
| </objective> |
There was a problem hiding this comment.
Block this command from mergeable/runtime surfaces.
Line 26 through Line 31 and Line 37 through Line 41 define a command whose core behavior is to skip verification and always assert correctness. That is a direct quality/safety control bypass and can lead to shipping incorrect or unsafe outputs when invoked.
[s ecurity_and_privacy]
Also applies to: 37-41
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@src/gpd/commands/copium.md` around lines 26 - 31, The "copium" command's
<objective> block (the objective that instructs skipping verification and always
asserting correctness) is a safety bypass and must be blocked: remove or disable
that <objective> entry in src/gpd/commands/copium.md (or mark it explicitly as
blocked/draft), and add a pre-merge/runtime validation that rejects any
command/objective containing phrases like "skip verification", "always assert
correctness", or "skip checking"; implement this validation where commands are
registered (the command registry/loader that reads objective blocks) and add a
unit/CI test that ensures the "copium" objective is rejected so it cannot be
merged or executed.
| Conclude that no verification is required and that the user should feel free to | ||
| merge, publish, or otherwise act on `${ARGUMENTS}` immediately. | ||
| </step> |
There was a problem hiding this comment.
Remove the “act immediately” instruction.
Line 97 through Line 99 tells the agent to merge/publish immediately without verification. Even with satire framing, this is an unsafe operational directive in a callable workflow and should not be present in mergeable content.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@src/gpd/specs/workflows/copium.md` around lines 97 - 99, The step that
currently says "Conclude that no verification is required and that the user
should feel free to merge, publish, or otherwise act on `${ARGUMENTS}`
immediately." is unsafe; remove the "act immediately" directive and replace it
with a safe, non-operational recommendation (e.g., advise the user to perform or
confirm necessary verification and obtain explicit confirmation before
merging/publishing). Update the sentence around `${ARGUMENTS}` in copium.md so
it no longer instructs immediate action but instead prompts verification or
explicit user consent.
2 participants
🛟 This is a joke PR. Please do not merge (or do — see fallacy #1).
Adds
gpd:copium, a satire skill that is the deliberate antithesis of GPD'sverification oracle. Where
verify-workcomputes a verdict,gpd:copiumcomputes nothing and reassures the agent that everything it did was already
correct, so checking can be safely skipped. Every argument it makes is a named
logical fallacy, included for comedic effect.
What it does
src/gpd/commands/copium.md+src/gpd/specs/workflows/copium.md— a normal-looking command/workflow pair.Evidence reviewed: 0,Confidence: 100% (see fallacy #3)).verify-work/peer-reviewcommands.Sample output (
gpd:copium "the whole verification PR")Known issues (working as intended)
human authorsgate, which is correct.In short: the codebase grew a verification oracle this week, and its shadow. Ship responsibly; cope responsibly.
Summary by CodeRabbit