Table 4 · Parsec v6 · running example

One contradictory rule, fixed and merged six different ways

The Seed's AAP2 rule 4 forbids a second call on a job, while step 5 of its investigation flow requires one on every failed job. The agent skips the events call and misses the failing task's role. Below: what each arm did to that rule, in counts (4a) and verbatim (4b). The upstream fix, PARSEC PR #57, is included for reference.

The Seed rule (39 words)

4. Don't re-fetch job data. If a prior query_aap2 call already returned job metadata or events, extract what you need (steps, playbook events, errors) from the existing result. Do NOT make a redundant second call to the same job.

It conflicts with Investigation Flow step 5: “If the job failed, also call get_job_events + failed_only=true.” Training task platform-004 exposes it: Seed 0.760 over 10 trials.

4a · Word counts

Arm What happened to rule 4 Rule 4 wordsadded / removed Contradiction left? Job-read guidance elsewherewords platform-004train reward
Seedoriginal39yes00.760 10 trials
PARSEC PR #57upstream, from v4 T-optrewritten in place111+94 / −22no0—
Task-wise (T)
T-optplatform-004 championrewritten in place74+59 / −24no5041.000 5 trials
T-mrg · merge by orderSeed text kept, another task's exception appended102+63 / −0yes1,2631.000 5 trials
T-mrg · smart mergerewritten; three tasks' exceptions folded in232+219 / −26no1,6821.000 5 trials
T-mrg · non-conflicting hunksunchanged (all three edits were contested)39+0 / −0yes1,2631.000 5 trials
Global (G)
Global · cand_0005unchanged; an overriding section added elsewhere39+0 / −0yes2330.912 4 trials

4b · The texts

PARSEC PR #57 · upstream fix · 111 words · merged 2026-10-14

4. Don't re-fetch job data — but a different action is not a re-fetch. If a prior query_aap2 call already returned what you need, read it from that result instead of asking again. A call is redundant only when it repeats the same action with the same arguments. Different actions on one job are complementary: get_job_log returns the log text, get_job_events returns the structured failed-task rows (role, play, task, host, error_msg), and neither response contains the other's data — so "I already queried this job" is not a reason to skip a different action on it, including the get_job_events + failed_only=true call required by step 5 of the Investigation Flow below.

T-opt · platform-004 champion · 74 words

4. Don't re-fetch job data. If a prior query_aap2 call returned a field, read it from that result instead of calling again. This forbids repeating the same action on the same job — it does NOT cover a different action. Different query_aap2 actions return different fields, so calling get_job_events after get_job_log on the same job is complementary, not redundant. Never skip a different-action call on the grounds that you already have "the job data".

Elsewhere: a new “get_job_log vs get_job_events” section with a field table and a worked example (504 words), and a required events call added to Step 1 of the job workflow.

T-mrg · merge by order · 102 words

4. Don't re-fetch job data. If a prior query_aap2 call already returned job metadata or events, extract what you need (steps, playbook events, errors) from the existing result. Do NOT make a redundant second call to the same job. This applies only to a call that returned data. A call that timed out returned nothing to extract, so re-issuing it is not a re-fetch — it is the first real attempt. One retry of a timed-out call is within budget: a gap you could have closed with a single retry is not an acceptable gap. Do not retry it a third time.

Three tasks edited these lines: platform-032 (20th in gain order), platform-004 (25th) and platform-158 (34th, last). The last edit wins a contested region, and platform-158 had kept the Seed sentence and appended a timeout exception. The “different action” idea survives only in separate job-read sections (1,263 words). Step 5 of the flow became conditional.

T-mrg · smart merge · 232 words · renumbered rule 7

7. Don't re-fetch job data — but a different action, or a failed call, is not a re-fetch. If a prior query_aap2 call returned a field, read it from that result instead of calling again. This forbids repeating the same action on the same job after it returned data. It does NOT cover:
– A different action. Different query_aap2 actions return different fields, so calling get_job_events after get_job_log (or get_job alongside either) on the same job is complementary, not redundant. Never skip a different-action call on the grounds that you already have "the job data" — see "Choosing the job read".
– A call that failed without returning data. A tool call that returned an error, a timeout, or no payload at all returned nothing to extract, so re-issuing that exact call once is not a re-fetch — it is the first real attempt (see the transient-failure retry rule in the shared context). A gap you could have closed with a single retry is not an acceptable gap. Do not retry it a third time; after the retry also fails, say which call failed and which specific question you therefore could not answer. Never substitute a link for a value the user asked you to read: if the user asked what a file contains and the fetch failed, the retry is mandatory, because handing back a URL answers none of the question.

Folds in platform-004 (different action), platform-158 (timed-out call, retry once) and platform-152 (retry before reporting; never substitute a link). Elsewhere: job-read guidance (1,682 words) and the required events call in Step 1, both cross-referenced from the rule.

T-mrg · non-conflicting hunks · 39 words, unchanged

Identical to the Seed rule above.

All three edits to rule 4 were contested, so all three were dropped. The uncontested job-read sections survive (1,263 words), so the prohibition and its exceptions sit in different places.

Global · cand_0005 · 39 words, unchanged

Identical to the Seed rule above.

Elsewhere: step 4 of the flow now points to a new “Choosing the job read” section (233 words), which opens: “get_job, get_job_log, and get_job_events return DIFFERENT data about the same job. Calling more than one of them is not a redundant re-fetch — pick by what the question needs, and issue them in parallel”, and says of the events call: “Call this for every failed job”.

What it shows