Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/XxYouDeaDPunKxX/PA-PVP-mini/llms.txt

Use this file to discover all available pages before exploring further.

Every PA-PVP Mini response follows a fixed, fully structured format. The output is designed to be machine-parseable and unambiguous — field positions are consistent, labels are from a closed vocabulary, and the END OF PROTOCOL marker signals clean termination. No prose, narrative, or commentary appears outside the defined fields.

​
ROUND Header

The output opens with a ROUND block that identifies the current review round and the mode of execution.
FieldTypeValues
current_roundnumberThe round number in effect for this output
round_modestringROUND 1 or ROUND 2+

​
FINDINGS Block

Each finding occupies one block. Fields appear in the order shown below. All fields are required for every finding.
​
finding_id
string
required
A single lowercase keyword taken from the core problem — for example padding, merge-conflict, or sev-boundary. No numbers, no category prefixes. Once assigned, this ID must be reused exactly across rounds whenever the same problem recurs. Do not assign a new ID to a problem already tracked in PREVIOUS_ROUND_OUTPUT.
​
fix_id
string
required
Same naming convention as finding_id. Identifies the fix associated with this finding. Used in the MERGE STATUS block to track fix disposition across rounds.
​
origin_round
number
required
The round number in which this finding first appeared.
​
derived_from
string
required
The finding_id of a prior finding that this finding replaces or evolves from. Use NONE if this finding is not derived from any prior finding.
​
severity
string
required
HIGH, MED, or LOW. See Severity Levels for definitions.
​
label
string
required
INDEPENDENT in Round 1 for all findings. In Round 2+: INDEPENDENT (produced without consulting prior output), CONFIRMED (prior finding revalidated), CHALLENGED (prior finding rejected, overstated, duplicated, or weak), or NEW (not present in prior output).
​
what_breaks
string
required
One concise statement describing what structural element fails or is absent.
​
evidence
string
required
The observable reason anchored to ORIGINAL_ARTIFACT. Must cite specific content from that block.
​
fix
string
required
The structural fix. Must change at least one of: structure, behavior, constraints, decision rules, flow, validation, or operational clarity. Pure wording fixes are excluded unless they resolve structural ambiguity.

​
MERGE STATUS Block

The MERGE STATUS block appears only in Round 2+ output. It summarizes the disposition of all findings and fixes relative to prior rounds. It has four fields.
​
confirmed_fixes_ready
list
Each entry: [fix_id] + a one-line reason the fix is confirmed as valid and ready to apply.
​
challenged_items
list
Each entry: [finding_id or fix_id] + why it is challenged + replacement finding or fix if one exists. Challenged items are those the independent review found to be invalid, overstated, duplicated, or structurally weak.
​
new_fixes
list
Each entry: [fix_id] + a one-line reason this fix is new (no prior counterpart).
​
conflicts
list
Each entry: [finding_id] + the independent review’s position + the prior audit’s position. A conflict exists when the two rounds reach opposite conclusions about the same problem.

​
END OF PROTOCOL Marker

The literal string END OF PROTOCOL closes every output. It signals that the structured output is complete and no additional content follows.

​
Example Outputs

Round 1 example: 
ROUND:
- current_round: 1
- round_mode: ROUND 1

FINDINGS:
- [null-check] [null-check-fix] [1] [NONE] [HIGH] [INDEPENDENT]
- what_breaks: The validation step has no defined behavior when input is null.
- evidence: ORIGINAL_ARTIFACT section "Input Handling" specifies required fields but contains no null branch.
- fix: Add an explicit null check before the validation block that returns a typed error and halts execution.

- [retry-cap] [retry-cap-fix] [1] [NONE] [MED] [INDEPENDENT]
- what_breaks: The retry loop has no maximum iteration count, allowing unbounded execution.
- evidence: ORIGINAL_ARTIFACT section "Retry Logic" defines a delay interval but no termination condition.
- fix: Add a configurable MAX_RETRIES constant and a loop exit condition that raises a failure state when exceeded.

END OF PROTOCOL
Round 2 example: 
ROUND:
- current_round: 2
- round_mode: ROUND 2+

FINDINGS:
- [null-check] [null-check-fix] [1] [NONE] [HIGH] [CONFIRMED]
- what_breaks: The validation step has no defined behavior when input is null.
- evidence: ORIGINAL_ARTIFACT section "Input Handling" still contains no null branch after revision.
- fix: Add an explicit null check before the validation block that returns a typed error and halts execution.

- [retry-cap] [retry-cap-fix] [1] [NONE] [MED] [CHALLENGED]
- what_breaks: Prior finding claimed unbounded retry loop; revised artifact now includes MAX_RETRIES.
- evidence: ORIGINAL_ARTIFACT section "Retry Logic" now defines MAX_RETRIES = 5 with an exit condition.
- fix: No fix required — prior finding is resolved.

- [fallback-state] [fallback-state-fix] [2] [NONE] [MED] [NEW]
- what_breaks: Failure state after MAX_RETRIES is reached has no defined fallback behavior.
- evidence: ORIGINAL_ARTIFACT section "Retry Logic" exits the loop on MAX_RETRIES but specifies no subsequent action.
- fix: Define a fallback path — either a dead-letter queue entry, a caller-facing error response, or an explicit halt — that executes after the retry cap is reached.

MERGE STATUS:
- confirmed_fixes_ready: [null-check-fix] Still unresolved in revised artifact; fix remains valid and applicable.
- challenged_items: [retry-cap-fix] Resolved in revised artifact; prior finding no longer applies. No replacement.
- new_fixes: [fallback-state-fix] New gap exposed by the retry cap addition in the revised artifact.
- conflicts: NONE

END OF PROTOCOL

Build docs developers (and LLMs) love

Get started for freeTalk to us