e10-rtm-drd
Use when build or audit a requirements traceability matrix against the ECSS-E-ST-10C Annex N Document Requirements Definition: confirm every row carries its DRD fields, trace each derived requirement
SKILL.mdskills/space-systems/ecss/e10-rtm-drd/SKILL.md
GitHubECSS System Engineering — Requirements Traceability Matrix DRD (space-systems/ecss/e10-rtm-drd)
Use when the task is to build or audit a Requirements Traceability Matrix against the Document Requirements Definition of ECSS-E-ST-10C Annex N -- the per-row trace fields, and the two directions of coverage the matrix exists to prove.
Domain quick reference
- Traceability runs in two directions and each finds a different defect. Upward: a derived requirement that reaches no parent came from nowhere. Downward: a requirement that must flow down but has no child was accepted and never allocated onward. Neither check substitutes for the other.
- A top-level requirement is declared, not inferred from a blank parent field. That distinction is the point: if absence of a parent meant "root", every forgotten link would silently become a new root and the upward check would find nothing.
- The two declarations are cross-checked both ways. A row declared top-level that also names a parent is as much a finding as a derived row with no parent -- the matrix is contradicting itself.
- A parent that is not itself a row in the matrix is a dangling trace: it looks traced, and it leads out of the document.
- Parent pointers can close into a cycle, which makes the upward walk non-terminating. Cycles are found explicitly rather than by letting a traversal run away.
- Allocation is checked against the product tree, not merely for presence. A row allocated to an element that does not exist is untraceable in exactly the way the matrix is meant to prevent.
- A blank verification method and an unrecognized one are separate findings, raised by separate checks, so one row never double-reports.
Workflow
- Confirm every row has a requirement identifier and that identifiers are unique; stop if not, since an ambiguous key makes tracing meaningless.
- Check each row for its required DRD fields.
- Run the upward trace: top-level rows must name no parent, derived rows must name one, and every named parent must exist.
- Run the downward trace: each row expecting flow-down must have at least one child.
- Detect parent-pointer cycles and report every requirement on one.
- Check each allocation against the product-tree element set, and each verification method against the recognized four.
- The matrix is Annex N complete only when the finding list is empty.
Pitfalls
- Inferring "top level" from a blank parent field, which converts every dropped link into a spurious root and empties the upward check of meaning.
- Checking the upward trace only. Coverage of the parent requirements is the half that proves nothing was accepted and forgotten.
- Walking parent pointers without a cycle guard, which hangs on exactly the malformed matrix the audit is supposed to catch.
- Accepting an allocation string because it is non-empty. A typo names a product element that does not exist and still looks allocated.
- Reporting a blank verification method twice, once as a missing field and once as an unrecognized method. It is one defect.
- Reading a high percentage of traced rows as a pass. Annex N coverage is all-or-nothing; the untraced remainder is where the risk sits.
Behavior contract (gate 3)
The row-uniqueness, DRD-field, upward-trace, downward-trace, cycle-detection, allocation and verification-method logic is exercised by the gate 3 contract test: scripts/test_e10_rtm_drd.py against scripts/e10_rtm_drd_logic.py (stdlib unittest, offline). Run: python3 scripts/test_e10_rtm_drd.py
Compliance
- ECSS standards are freely downloadable (ESA); cite the source and paraphrase per standards-map.yaml.
- compliance: STANDARDS-REF, gated: false.