Paper Policy
Assess manuscript integrity, evidence and applicable project or venue requirements.
Editorial judgment lives in the writing, review and artifact guides. Choose wording,
structure and presentation for the task without preference-status records.
Read references/authority-model.md when resolving conflicts. Reliable venue or
project requirements can determine format; they cannot authorize fabricated facts
or unsupported claims. Keep artifact creation in its owning skill and evidence
management in the user-selected workflow.
Choose the Operation
- Ordinary writing or critique: apply relevant integrity and explicit project
constraints through the owning skill. No policy CLI, context file or full
readiness assessment is required for a section edit.
- Applicability or rule conflict: read
references/context-schema.md and
explain which requirements apply and why.
- Compliance or submission readiness: read the context and compliance schemas,
run relevant deterministic checks and inspect actual evidence.
- Skill maintenance: use
references/rule-maintenance.md and validate changed
registries, consumers and references together.
An existing context file or final-stage label does not expand the requested scope.
Briefs and working notes support continuity when useful; source snapshots needed
for an evidence-backed compliance PASS remain mandatory.
Resolve and Assess
references/hard-rules.yaml contains manuscript requirements.
references/profiles.yaml activates conditional requirements from task context.
The scripts locate these files relative to the installed skill, with manuscript
paths supplied explicitly. Read references/constraint-schema.md for maintenance.
python3 /path/to/skills/paper-policy/scripts/resolve_policy.py /path/to/paper_context.yaml
python3 /path/to/skills/paper-policy/scripts/run_project_validation.py \
/path/to/paper_context.yaml /path/to/paper --output-dir /path/to/validation-output
python3 /path/to/skills/paper-policy/scripts/assess_compliance.py \
/path/to/paper_context.yaml --project /path/to/paper \
--evidence /path/to/compliance-evidence.yaml
- Inspect scope, warnings and context provenance before interpreting results.
Inference cannot activate provenance-sensitive requirements.
- Use
primary_tex and optional additional_tex for multiple document roots;
audit their include trees and referenced bibliographies.
- Generated artifact skeletons and evidence worklists are inspection aids, not
evidence. Confirm discovered artifacts against their sources. Unused bibliography
keys are maintenance candidates, not permission to delete entries.
- Check type is
deterministic, semantic or manual. Missing admissible
evidence leaves a requirement UNVERIFIED. Agent semantic PASS needs artifact,
locator, reasoning and current source snapshots. Manual PASS requires human,
user or venue evidence. Deterministic FAIL takes precedence over supplied PASS;
agents cannot waive requirements. Follow references/compliance-schema.md.
- For a PDF or excerpt, inspect only what the available material supports and
report unavailable checks. An empty TeX project cannot establish readiness.
Output and Mutation
Resolution returns context, active requirements, activation reasons and unverified
inputs. It does not assign PASS. Assessment adds evidence-backed statuses and
unresolved checks. Report editorial recommendations only when useful to the task;
ordinary edits need no policy dump or per-choice explanation.
Assessment is read-only. When manuscript or bibliography fixes are already
requested, continue through the owning skill and make scoped, supported changes.
Registry validity does not establish scientific correctness or readiness.
Resources
references/authority-model.md: authority, waivers and execution boundaries.references/constraint-schema.md: requirement and activation-profile schema.references/context-schema.md: applicability, manuscript selection and provenance.references/compliance-schema.md: evidence, freshness and readiness semantics.references/paper-context.example.yaml: example formal-audit context.scripts/validate_registry.py: structural validation.scripts/audit_skill_integration.py: explicit references and registered rule IDs.scripts/check_artifacts.py: artifact sources, paths, formats and coverage.scripts/discover_artifacts.py: guarded LaTeX artifact inventory.scripts/project_files.py: selected TeX trees and bibliographies.scripts/lint_project.py: definite findings and contextual review hints.scripts/run_project_validation.py: resolution, lint, assessment, discovery,
evidence worklist and located unused-key report.